The Impact of the German Child Benefit on ChildWell-Being

Christian Raschke∗

October 2012

Abstract

The German Child Benefit (“Kindergeld”) is paid to legal guardians of childrenas a cash benefit. This study employs exogenous variations in the amount of childbenefit received by households to investigate the extent to which these various changeshave translated into an improvement in the circumstances of children related to theirwell-being. I use the German Socio-Economic Panel to estimate the impact of a givenchange in the child benefit on food expenditures of households, the probability ofowning a home, the size of the home, as well as the probability of parents’ smoking,alcohol consumption, and parents’ social activities such as traveling, visiting movietheaters, going to pop concerts, attending classical music concerts or other culturalevents. Households primarily increase per capita food expenditures in response toincreases in child benefit, and they also improve housing conditions. I do not finda significant effect of child benefit on parents’ smoking or drinking, but parents ofolder children use the child benefit to pay for their social and personal entertainmentactivities.

JEL Codes: I38, D12, H31.Keywords: Child Benefit, Fungibility of Income, Child Well-Being.

∗Department of Economics, Louisiana State University, 2305 Business Education Complex, Baton Rouge,LA 70803, USA. E-mail: crasch2@lsu.edu. I thank Naci Mocan as well as seminar participants at LSU andthe IZA Summer School 2012 for very constructive comments.

1

1 Introduction

The parents of children in Germany receive a monthly child benefit (“Kindergeld”) for every

child. This benefit is not means-tested and all German and EU citizens living in Germany

qualify for it.1 The Child Benefit is at the core of all family support policies in Germany

and the Federal Constitutional Court has confirmed that it is intended to ensure that chil-

dren’s needs are met in terms of nutrition, housing, clothing, personal hygiene, and heating

(Bundesverfassungsgericht, 1998). Since the benefit is not provided in-kind but rather as a

cash transfer to the parents, how much of the additional income is actually used to improve

the well-being of children remains an open empirical question.

While there exists some literature investigating the impact of child benefit systems in

different European countries and Canada, this study is the first to investigate the impact of

child benefit income on various dimensions of household expenditures and parental behavior

in Germany. More importantly, this paper contributes to the literature by cleanly identifying

the effect of the child benefit. I use a panel data set of households and exploit exogenous

intra-household variation in the amount of the benefit to isolate its effect on the outcomes of

interest. Using within-household variation of child benefit income in a panel of households

and controlling for various time-varying household characteristics overcomes the identifica-

tion problem of separating the effect of a child benefit from the effects of differences in the

number of children, ages of children, and other confounding factors noted in the previous

literature (Edmonds, 2002).

Standard economic theory of consumer choice predicts that the source of any additional

household income is irrelevant in determining how that additional income is spent. If families

maximize a joint utility function, then a cash transfer shifts out the family budget constraint

and the change in expenditures resulting from the shift in the budget constraint does not

1Parents who are German citizens but live abroad may be eligible if they earn income that is fullytaxable in Germany. Non-EU citizens may be eligible for child benefit depending on their immigrationstatus. The details of the eligibility rules can be found in §62 Einkommensteuergesetz (EStG) and §1Bundeskindergeldgesetz (BKGG).

2

depend on the source of the additional income. An increase in child benefit will result in

the same Engel curves as would an increase in other types of household income.2 However,

Thaler (1990) provides evidence suggesting that individuals do not treat cash income from

different sources as perfectly substitutable (also see Thaler, 1999). He argues that “mental

accounts” attach labels to income from different sources and that the marginal propensity

to consume may be different for each of the income sources.

The labeling effect described by Thaler (1990) has been studied empirically in the context

of child benefit. Kooreman (2000) used repeated cross sections of Dutch consumer expendi-

ture surveys to investigate the effect of the Dutch child benefit system on parents’ spending

behavior. He found that the effect of an increase in child benefit income on child clothing

expenditures is larger than the effect of an increase in income from other sources on child

clothing expenditures for households with only one child. Kooreman suggested that this may

be caused by a labeling effect, where parents keep “mental accounts” as proposed in Thaler

(1990). However, in Kooreman’s study this labeling effect vanished in households with more

than one child: larger Dutch families did not distinguish between child benefit income and

income from other sources. In a related study, Edmonds (2002) used cross-sectional house-

hold data of the Slovenian child benefit program from 1993 to investigate the impact of the

child benefit on household expenditures for food, tobacco, alcohol, and clothing. He found

no significant effect of child benefit income for any of the above expenditures and concluded

that there was no labeling effect.

Most recently, Blow et al. (2012) examined the United Kingdom’s Child Benefit program.

They studied the time period of 1980-2000 and relied on variation in the real child benefit

due to the fact that the government imperfectly accounted for inflation when adjusting

2Households do no necessarily have to make decisions as a unit for this result to hold: Becker (1981)describes a model in which one person in the household controls all household resources, but is altruistictoward other members of the family. In this case a different intra-household distribution of income does alsonot result in different consumption patterns for the household. On the other hand, the model developed byLundberg and Pollack (1993) predicts that cash transfer payments to the mother and payments to the fathercan imply different consumption patterns for the household (also see Browning et al., 1994). Lundberg et al.(1997) and Schady and Rosero (2008) provide empirical support for the theoretical results of Lundberg andPollack (1993).

3

the nominal child benefit, and that the amount of child benefit received for the first child

increased considerably on two occasions due to a policy change. They found that up to half

of a marginal dollar of an increase in the UK child benefit is spent on alcohol. Changes in

the UK’s Child Benefit policies have also been used by Lundberg et al. (1997) to show that

expenditure patterns are different when the child benefit is paid to the mother, compared to

when it is paid to the father. They show that expenditures on women’s and child clothing

increased when a reform of the UK’s system transferred the payout of the benefit from the

head of household to the mother.

Given these sharply different results, this paper adds to the evidence on the effects

of child benefit income. Data limitations prevent me from analyzing the impact of child

benefit on expenditures for child-specific goods, such as toys or child clothing. Instead,

I estimate a number of household-level equations explaining households’ choices regarding

food expenditures and housing. I also investigate individual-level equations to see how

parents’ consumption patterns change in response to changes in child benefit. In particular,

I investigate whether variations in child benefit impact parents’ consumption of cigarettes

and alcohol, parents’ personal entertainment activities, participation in cultural events, as

well as vacation travel. Some of these commodities are clearly unrelated to child well-being;

others could be related, as I explain later in this section. The results provide insights into

whether the child benefit is spent according to the intention of the policy maker.

There is a strong relationship between food expenditures and child well-being, as well as

between the housing arrangement of families and child well-being. Food expenditures are

directly related to food security (Nord et al., 2010).3 Furthermore, previous work has shown

that food expenditures are highly correlated with dietary quality. For example, Mabli et al.

(2010) found that food expenditures of low income households are positively related to the

proportion of fruit and vegetables in the households’ diet, nutrient density, and expenditure

shares of foods recommended for frequent consumption. The authors also found that energy

3This relationship can be considered mechanical: Food security arises due to a lack of resources (andspecifically due to a lack of money) to purchase food.

4

density and expenditure shares on foods that are not recommended for frequent consumption

decrease in food expenditures. Since the well-being of children improves when they are eating

a healthier diet at home, it is important to investigate how child benefit income affects per

person food expenditures in the household.

Regarding housing outcomes, Green and White (1997) found substantial effects of home

ownership on child outcomes. In particular, they found that residential stability associated

with home ownership is correlated with higher school attainment of children, keeping family

income constant. Aaronson (2000) confirms this finding after accounting for more detailed

observable family characteristics and attempting to control for endogeneity of homeowner-

ship. In addition to ownership, the size of the home also plays a role in child well-being.

Crowding occurs when the physical size of the home is too small for the number of family

members, and Goux and Maurin (2005) find a negative causal effect of overcrowded housing

conditions on academic performance of children. I estimate the impact of child benefit on

the probability of home ownership, as well as the size and number of rooms in a home.

Since the child benefit is fungible, parents may use child benefit to support their con-

sumption of cigarettes and alcohol, or other commodities that are unrelated to the well-being

of children. It is not the intention of the policy maker to encourage parents to smoke or

drink alcohol, and therefore I also estimate a number of individual-level equations explain-

ing parents’ smoking and drinking behavior. The literature on the economics of smoking

finds the income elasticity of smoking cigarettes varies systematically over time and across

countries (Cheng and Kenkel, 2010; Wilkins et al., 2001). There is a strong income gradient

in cross sections indicating that lower income individuals are more likely to smoke in the

United States, as well as in Germany (Chaloupka and Warner, 2000; Nocon et al., 2007).

At the same time, recessions (i.e. time periods when average incomes are low) are associ-

ated with improvements in health behaviors such as smoking and drinking (Ruhm, 2000,

2005), Chaloupka and Grossman (1997) showed that income is positively associated with

smoking for youths, and Kenkel et al. (2011) found that smoking is a normal good for low

5

income households with children. In order to test whether child benefit contributes to par-

ents smoking or drinking, I estimate the impact of child benefit on smoking participation, the

number of cigarettes smoked (conditional on smoking), as well as the probability of regularly

consuming alcohol.

In addition to the health behaviors related to smoking and drinking, I test whether

parents use the benefit to pay for their social activities and personal entertainment activities.

These activities include visits to the movie theater or pop music concerts, attending the

opera, public lectures, or participating in other cultural events, as well as making excursions

or taking short vacation trips. The personal entertainment activities of the parents, such as

going to the movies or pop music concerts are unrelated to child well-being. On the other

hand, making excursions or going on day trips may include trips to the zoo or amusement

parks. Therefore, this activity may be positively related to the well-being of children.

Using data from the German Socio-Economic Panel from 1998 to 2009, I find that house-

holds largely use the child benefit as intended by the policy maker. Households respond to

increases in child benefit income by increasing per capita food expenditures. Specifically,

households spend between 49 and 74 cents out of each additional Euro of child benefit on

food. Outcomes related to housing also improve, especially for parents with younger chil-

dren. Additional child benefit income is associated with a marginally larger home, and an

increased likelihood to own a home instead of renting. I do not find any evidence that par-

ents increase smoking or drinking alcohol in response to increases in child benefit. However,

parents of older children (at least 18 years old) who still receive child benefit are more likely

to use child benefit for their own social activities and are more likely to attend pop concerts,

go to the movies, or go to the opera/theater at least once a month.

I also find evidence for a significant labeling effect. Households spend more of every

marginal Euro of child benefit on improving nutritional quality and on improving housing

conditions, in comparison to an additional Euro from other sources of income.

6

2 Background of the Child Benefit in Germany

The basic idea of the child benefit in Germany is as follows:4 The living wage of an individual

or couple is not subject to income tax and only income that exceeds the level of living wage

(“Grundfreibetrag”) is taxable.5 Households with children are eligible for an additional tax

deduction for each child (“Kinderfreibetrag”), but this additional deduction is not considered

when calculating tax withholdings. Instead, families with children receive a monthly child

benefit in cash. The cash benefit is strictly a function of the number of children; that is,

household income or other considerations are irrelevant. All children are generally eligible

until they turn 18 years old. Thereafter, the parents may still receive child benefit for the

child until the child turns 25 if certain conditions regarding employment status and earnings

of the child are met. For instance, a family may continue to receive child benefit for a child

older than 18 years if the child still attends school or university. 6 Conditional on the number

of children, the child benefit amount is exogenous and does not depend on any other family

characteristics. This means that a household with two children pays the same amount of

income tax as a childless but otherwise identical household with identical income; yet, the

household with children receives the cash child benefit while the childless household does

not.

At the end of the year, the Tax Office calculates the total yearly amount of cash child

benefit received by the household, and it also calculates the potential tax savings if the child

deduction would have been taken instead. For most households the total yearly cash child

benefit they received throughout the year exceeds the tax savings that would have resulted

from claiming the child deduction. In this case the household keeps the excess amount as a

transfer payment (“Forderanteil”). For households with very high income, the tax savings

4In this section I describe the child benefit system in place during 1996–2011. See Ruhl (1994) for aninteresting historical perspective on different approaches to family policy in Germany starting with the early1900s.

5This is equivalent to a “personal exemption” in the U.S. income tax system.6The precise conditions to be able to receive the benefit beyond the 18th birthday of the child changed

over time and are not considered in this paper.

7

from taking the child deduction exceeds the total yearly child benefit. In this case, the

household claims the deduction and pays back any child benefit received.7 The Tax Office

determines ex officio whether the child benefit or the child deduction results in larger savings

for the household and therefore this choice is not subject to household decision or to errors

by the filing household. In monetary terms, direct government expenditures due to the

child benefit and the child deduction were e38.5 billion in 2009 (about 1.6% of GNI), e36.9

billion of which were for the cash benefit and the remaining 1.7 billion for the tax deduction

(Bundesministerium der Finanzen, 2010).

Taxable income is not directly observed in the data and is difficult to accurately impute.

Moreover, due to changes in the tax schedule and changes in the amount of the allowed

deduction, the taxable household income threshold, beyond which taking the tax deduction

provides a larger benefit in comparison to taking the cash payment, varies slightly over time.

Therefore, in this paper I use only the cash amount of child benefit that is paid out to

everyone throughout the year. Ignoring the possibility that households with high income

may have received an additional benefit when filing their tax return means that the amount

of cash child benefit paid throughout the year can be considered a lower bound. In 2009,

there were 17.7 million children eligible for child benefit in Germany and only 11.8% of those

children lived in households with sufficiently high household income to be considered for the

tax deduction instead of the cash benefit (Bundesministerium der Finanzen, 2010). Since

the majority of households receive the cash benefit, I do not expect that using the lower

bound of benefit received will have an impact on the results presented below.

7The current system of receiving a large cash benefit, or a tax deduction has been in place since 1996.Prior to 1996, households received both a tax credit and a (much smaller) cash benefit payment. Previousstudies have used changes in the child benefit law as an exogenous source of income variation. For example,Tamm (2008) used the 1996 change in the law as a “natural experiment” that increased income for householdswith children. He used this natural experiment to determine that income is not causally related to schoolchoice in Germany.

8

3 Empirical Strategy and Data

I estimate household-level specifications as well as individual-level specifications. The out-

comes of interest used in this paper that are observed at the household level are real per

capita monthly food expenditures of the household, the probability that the household rents

their home, the size of the home measured in square meters, the number of rooms in the

home, and the amount of rent paid per month for renters. Smoking and drinking relates to

the behavior of individual persons and therefore the unit of observation is the individual par-

ent. I estimate equations describing the probability that a parent smokes, the daily number

of cigarettes that a parent consumes conditional on smoking, as well as the probability that

a parent regularly consumes alcohol. I also estimate at the individual level the probability

that parents attend pop concerts or go to the movies at least once a month, the probability

that parents go to classical concerts or attend other cultural events at least once a month,

and the probability that parents go on excursions or short vacation trips at least once a

month.

Let the outcome of interest k of household j during time t be described by the equation

Outcomejtk = fk (Yjt; Zjt) , (1)

where Yjt is the real net monthly household income of household j during year t and Zjt

contains other exogenous time-varying characteristics of the household. If real monthly net

household income consists of the child benefit, yjt, and income from other sources, mjt, then

Outcomejtk = fk (yjt,mjt; Zjt) . (2)

Kooreman (2000), Edmonds (2002), and Blow et al. (2012) assume that fk(·) is linear, and

that y and m are additively separable. This means that these papers test for a labeling effect

of child benefit by comparing the coefficients of y and m in a linear regression of expenditures

on y, m, and control variables.

9

Using a comparison of the coefficients of y and m from a linear regression to identify

a labeling effect relies on the assumption that child benefit income and income from other

sources are additively separable. This is a strong assumption that deserves careful attention.

Child benefit income increases households’ unearned income and therefore may have an

impact on labor supply decisions. Theoretically, increases in unearned income result in a

decrease in labor supply and there have been several studies that lend empirical support to

this theoretical prediction specifically in the context of child benefit. For example, Tamm

(2009) provided evidence to suggest that the reform of the German child benefit system

resulting in a substantial increase of child benefit payments from 1995 to 1996 reduced the

labor supply of single mothers and mothers with a working spouse. Analyzing the data

from the Canadian child benefit for 1947-1999, McNown and Ridao-Cano (2004) found that

increases in child benefit reduced female labor supply in Canada.8

Out of the previous literature on the impact of child benefit discussed above, only Koore-

man (2000) explicitly discusses labor supply concerns. Motivated by the work of Browning

and Meghir (1991) who investigated the separability of commodity demands from labor

supply, Kooreman (2000) checked the robustness of his results by estimating expenditure

equations that include as an explanatory variable a dummy for mothers’ labor market par-

ticipation, instrumented by mothers’ education level9. However, the choice of the instrument

is crucial in order to identify the correct mechanism through which household income may

change (Moffitt, 2005). The concern is that child benefit acts as a budget shifter: keeping

constant wage, it can affect labor supply due to a shift in unearned income. Under this

8Researchers who are not specifically concerned with analyzing the effects of child benefit have alsorecognized that the labor supply effect of child benefit is important. For example, Blundell and Hoynes(2004) investigated whether a shift of the United Kingdom’s welfare policy toward in-work benefits between1980 and 2000 had a positive effect on labor markets. The authors recognized that an increase in theuniversal child benefit over the same time period may have offset some of the positive impact of the shifttoward in-work benefits on labor supply.

9Edmonds (2002) also instrumented other income, although his primary concern was measurement errorin household income. He used as instruments an indicator for whether or not the household is active inthe informal sector, regular employment, entrepreneurial activities, receiving fixed incomes, and numberof persons eligible for the child benefit. While Blow et al. (2012) wrote about comparing the marginalpropensities to spend out of child benefit and other income, they actually used total expenditure less childbenefit rather than income in their estimations.

10

scenario, an increase in child benefit can decrease labor supply, which in turn can lower

household income. Since child benefit is determined strictly by the number of children in

the household, it is orthogonal to any other household characteristics, including mothers’

education level. Even if education is a valid instrument for household income, instrumenting

household income with mothers’ education does not address the fact household income may

change due to a decrease in labor supply in response to an increase in child benefit.

Due to the fact that child benefit is exogenous to everything except for the number of

children in the household, finding a good instrument is challenging. In order to nonetheless

address any concerns regarding the potential labor supply effects of the variations in child

benefit, the main results presented in this paper are based on regressions that use the lagged

value of household income as the measure of household income. Child benefit can only affect

labor supply contemporaneously and in the future, and current year child benefit does not

influence labor supply decisions in the previous year. At the same time, household income is

expected to be highly correlated across years within the same household. Variations of this

specification are explored in the robustness checks section.

I estimate the following empirical counterpart of Equation (2)

Outcomejtk = αkyjt + βkmj(t−1) + Zjtδk + µjk + νsk + τstk + εjtk, (3)

where Zjt contains time varying observable characteristics of household j as well as controls

for local macroeconomic conditions captured by state level unemployment rates and an

indicator for whether Germany was in a recession during the time of the interview. µj is

a vector of household fixed effects so that the effect of child benefit can be identified from

variation in the child benefit within the households over time. νs is a set of indicators for the

state of residence of the household to account for systematic differences in behavior across

regional areas, and τst represents the set of region specific time trends to capture systematic

variation over time in each region. εjtk is an idiosyncratic error term.

11

When the outcome of interest is observed at the individual level instead of the household

level (e.g. for equations describing smoking participation, number of cigarettes smoked, and

the probability of regularly consuming alcohol) then Equation (3) becomes

Outcomeitk = αkyjt + βkmj(t−1) + Zijtδk + µik + νsk + τstk + εitk, (4)

where the subscript i denotes the individual. Note that in this individual-level fixed effects

specification the income variables mjt and yjt are still observed at the household level, and

Zijt now also contains time-varying exogenous characteristics of the individual in addition

to the household-level characteristics used in the specification of equation (3).

The main results of this paper are obtained by estimating specifications according to

Equation (3) and Equation (4).10 To address concerns regarding the impact of child ben-

efit on fertility, I estimate specifications separately for households that have one child and

households that have two children. Households that have one child and households with

two children account for 85 percent of the observations in my sample and therefore estimat-

ing specifications separately for households with three children or households with four or

more children would result in insufficient sample sizes. Therefore, in addition to estimating

specifications separately for households with one child and households with two children,

I estimate specifications using data on households with a varying number of children over

time. In those specifications I include a control variable for the number of children in the

household.

I also estimate specifications separately for families that have young children and families

that have older children. Families may receive the child benefit until the child is 25 years of

age. I define a household with young children to be a household in which the youngest child

is less that 18 years old. Equivalently, a household with old children is a household whose

children are all at least 18 years of age. Distinguishing between households with old children

10I will also report results of regressions that include contemporaneous child benefit and household incomeas a robustness check.

12

and households with young children is interesting for two reasons. First, the child benefit is

meant to ease the financial burden of raising children and to ensure that the needs of children

in terms of nutrition and housing are met. Younger children will have different needs in that

regard compared to older children. For example, older children are more likely to live outside

of the home of the parents and no longer eat their meals at home.11 Second, in 2012 a change

in the child benefit law expanded the eligibility criteria for children over the age of 18, while

not affecting younger children.12 If parents of older children behave systematically different

from parents of younger children in terms of how child benefit is spent, then identifying this

systematically different behavior will be important in the context of the new child benefit

law.

Data

I use data from the German Socio-Economic Panel (SOEP) to estimate specifications ac-

cording to equation (3) and (4). The SOEP is a longitudinal data set that has been in place

since 1984, with additional samples taken in 1993, 1998, 2000, and 2006 to counter sample

attrition.13 Among many other detailed household characteristics and individual character-

istics of each household member, the data also contain information regarding the amount of

child benefit received, as well as the number of children in the household (see Wagner et al.,

2007).14

The child benefit in Germany has been administered in its current form since 1996, and

11During the sample period used in this study, a child that is older than 18 but less than 25 years of agemay be eligible for child benefit while the child continues her education and the child’s own income fallsbelow a threshold level

12Specifically, the requirement that a child’s own income falls below a certain threshold has been removed.Since the change of the law took effect in 2012, it cannot be considered in this study and will be the subjectof future research.

13In addition, in 2002 there was a separate sample taken that targeted high income households. Due tothe fact that the tax savings due to a tax credit may be larger than the sum of the monthly cash child benefitpayments for this group, I exclude the 2002 high income sample from my estimations. My results are notsensitive to omitting or including this sample in the estimations.

14I use the official monthly amounts of child benefit for which households are eligible to correct for mea-surement error in the reported child benefit variable. Respondents are much less likely to commit an errorwhen counting their children compared to remembering the precise amount of child benefit they receive.

13

therefore the sample period used in this paper is very similar to the time period during which

the current child benefit policy has been in effect.15 Some of the outcomes that I consider in

this paper are not available in all years of the survey and I will use only data from the years

1998-2009 in this paper due to data availability. While households were asked about their

housing situation in every survey year 1998-2009, data regarding food expenditures have

only been collected in the years 1998, 2000, 2001, 2003, 2005, and 2007. Parents were asked

about their smoking behavior in 1999, 2001, 2002, 2004, 2006, and 2008, and about their

drinking behavior in 2006 and 2008. Individuals indicated whether they go to the movies

or to pop music concerts at least once a month and whether they attend cultural events at

least once a month during the survey years 1998, 1999, 2001, 2003, 2005, and 2007–2009.

Information on excursions and short vacation trips is available for the survey years 1998,

2003, and 2008. I do not hold my sample constant, but estimate all specifications using all

available data.

Table 1 presents summary statistics and description of the variables used in this study.

Panel (a) summarizes household-level variables and Panel (b) summarizes variables observed

at the individual level. I use only data on households with children and I exclude families

that have multiple generations living in one household. The average child benefit received

by households is substantial. Families receive on average 261.79 Euro per month in child

benefit for an average of 1.73 children per household. This means that on average, child

benefit income is equivalent to about 10 percent of other household income. The youngest

child in households is on average just over ten years old and single parents make up about

14 percent of the sample. As shown in Panel (b) of Table 1, parents are on average about

41 years old and about 35 percent of parents are smokers.

Figure 1 shows that there is variation in the nominal amount and real amount of child

benefit received over time, conditional on the number of children.16 In terms of real 2005

15There was a change in 1996 in the way that the child benefit was administered in Germany. Prior to1996, households received both a tax credit and a small cash benefit. From 1996 onward, the child benefitsystem has been administered according to the description in Section 2.

16The nominal changes of child benefit always took effect on January first of a given year.

14

Euros, a family with two children would receive e246.42 per month in 1998. By 2009, the

amount of the child benefit increased by about 25% to e306.54 per month for exactly the

same two children. Note also, that in real terms child benefit is not strictly increasing over

time. The amount of real child benefit was actually highest in 2002.

It is important to re-emphasize that, conditional on the number of children, the change

in child benefit is exogenous and does not depend on any other family characteristics. Issues

regarding the take-up of social assistance may be a concern if, for instance, a welfare stigma

discourages households to apply for support for which they may be eligible (Moffitt, 1983).

In the context of the German child benefit these issues are irrelevant. Due to the universal

eligibility of children, the child benefit is not considered social assistance in the same sense

that welfare programs are considered social assistance, and it is not associated with welfare

stigma (Frick and Groh-Samberg, 2007).

4 Main Results and Discussion

Food Expenditures

The first set of results pertains to per capita food expenditures of households. To the extent

that per person monthly food expenditures of a household are correlated with the nutritional

quality of the food consumed by that household, a positive impact of the child benefit on

per capita food expenditures would indicate that the child benefit is used according to the

intentions of policy makers. Total monthly household expenditures on food are observed

in the survey years 1998, 2000, 2001, 2003, 2005, and 2007 and I construct a measure of

per person food expenditures at the household level by dividing the total real monthly food

expenditures of households by the number of household members.

Table 2 shows that per capita food expenditures increase significantly as child benefit

increases. Point estimates of the effect range from an increase of per capita food expenditures

by 40 cents per additional Euro of child benefit received for households with one child

15

(Column 1), to about 13 cents per additional Euro of child benefit for households with two

children (Column 2). Columns 3 – 6 report the impact of child benefit on per capita food

expenditures for families with a different number of children over time, controlling for the

number of children in the household. Point estimates range from an increase in per capita

food expenditures by 15 cents for families with one or two children, to an increase of 21 cents

for households with any number of children. These results are highly statistically significant,

as well as economically significant.

Table 2 also present the results of estimating the impact of child benefit on per capita

food expenditures separately for households that have young children (the youngest child is

under the age of age 18) and households that have older children (the youngest child is at

least 18 years old). The results for the entire sample are driven by households that have

young children present: Regardless of the number of children in the household, the coefficient

of child benefit is statistically insignificant for households with older children. The fact that

child benefit has no effect on food expenditures in households with children older than 18

years of age is intuitive. Older children are less likely to eat at home because they are more

likely to attend college, participate in job training and eat their meals outside of the home.

The results in Table 2 show that household income also positively impacts per capita food

expenditures. However, the effect of child benefit income is significantly larger compared to

the effect of income from other sources, indicating that there is a labeling effect in child

benefit income with respect to food expenditures. An F-test rejects that the coefficient of

child benefit and the coefficient of other household income are identical with a p-value of

< 0.00. This is true for households with young children, as well as for the pooled sample

of households with children of any age. Since the effect of child benefit on per capita food

expenditures for households with older children is statistically insignificant, I do not find a

labeling effect for these households.

Multiplying the average household size by the estimated increase in per capita food

expenditures due to an increase in child benefit provides an estimate of how much total

16

household food expenditures change in response to changes in child benefit. Households with

only one child have 2.8 members on average. In this case the point estimate of the coefficient

of child benefit indicates that a one Euro increase in child benefit would result in an increase

of household food expenditures by more than one Euro. However, this point estimate is not

statistically significantly larger from the value that would suggest that all additional child

benefit is exhausted on food (p-value = 0.7141). Estimates for all other household sizes

are very reasonable and they range from an increase in total food expenditure of 49 cents

for every additional Euro of child benefit for households with one or two children, to and

increase of 74 cents per additional Euro for households with any number of children.

Housing

In this section I investigate the impact of child benefit income on outcomes related to housing.

In particular, I estimate the impact of the child benefit on the probability that the household

owns or rents their residence, as well as on the size of the residence.17 Questions about the

residential details of the household are asked in every wave of the survey so that changes in

the housing conditions of households can be tracked very precisely. Since most interviews in

the SOEP are completed in the first quarter of each year,18 I use the monthly amount of child

benefit received during the previous calendar year in all specifications related to housing.19

It is possible to purchase different food immediately in response to an income shock, but

housing decisions are subject to lease contracts and it may take some time to search for a

new home.

Table 3 shows that the probability of renting (versus owning) a home decreases in response

to an increase in the amount of child benefit. With the exception of the specification that

employs only households with two children (Column 2), this effect is statistically significantly

different from zero regardless of the number of children in the household. The magnitude of

17I exclude from the estimations households that live in public housing.18Two thirds of all interviews are completed by March. By April about 80% of all interviews are completed.19This also means that in order to account for the potential contemporaneous effect of child benefit on

labor supply, I include household income lagged by two periods in all housing related estimations.

17

the effect is also very consistent across family sizes: An increase in child benefit by about

100 Euro per month would be needed to increase the probability of home ownership by one

percentage point. Increases in household income also make it less likely that a household

rents their home, but the impact of an additional Euro in child benefit on the probability

of home ownership is larger than the impact of an additional Euro in household income.

Comparing the coefficient of child benefit and the coefficient of household income, an F-test

rejects the hypothesis that the coefficients are equal with a p-value of at most 0.052 for all

family sizes other than households with two children.20

The finding that increases in child benefit decrease the probability that household rent

their home is driven by households with younger children. The bottom panel of Table 3

shows that for households with older children, increases in child benefit have no statistically

significant effect on the probability of renting. At the same time, the impact of child benefit

on the probability of renting for households with young children is generally statistically

significant and of similar magnitude to the results for all households. It is intuitive that

child benefit does not influence the decision to rent or own in the case of households with

older children because children who are older than 18 years of age are very likely to move

out of the house soon.

Owning a home compared to renting a home can be considered an improvement in the

housing arrangement of a household, but home ownership by itself is not immediately related

to the well-being of children. The size of the home is much more important to determine

whether children are better off. Table 4 shows that a one Euro increase in child benefit is

associated with a larger apartment by 0.0095 square meters for households with one child.

This effect is statistically significant and the magnitude is consistent between specifications,

ranging from an increase of 0.0094 square meters per Euro for families with two or three

20In the interest of space, coefficients of control variables are not reported in the Tables, but are availableupon request. All of the control variables in this specification have the expected signs. For example, singleparents are more likely to rent their home, while households with a larger number of children are less likelyto be renters. while statistically indistinguishable from zero, the point estimates of the coefficient of thestate unemployment rate as well as the coefficient of the recession indicator are positive.

18

children, to an increase by 0.0145 square meters for households with one, two, or three

children. In terms of square feet, this means that the average home size increases by about

10 - 16 square feet for a 100 Euro increase in child benefit.

In addition to the overall home size, I also consider the number of rooms in the home. In

particular, if a child has a room to herself instead of sharing with siblings or other household

members, this is a direct improvement in housing conditions related to child well-being.

Table 5 present the impact of child benefit on the number of rooms larger than six square

meters in the homes of households. The effect of child benefit is positive and generally highly

statistically significant. The impact of child benefit on the number of rooms in the home is

statistically insignificant only for households with one child. On the other hand, households

with one child do increase the size of their home in square meters in response to increases in

child benefit, as shown in Table 4. This difference in results is consistent with the idea that

an additional room in the home will be more useful than additional square footage when

there are more children in the household.

The significant positive impacts of child benefit on home size (Table 4) and the number

of rooms in the home (Table 5) are driven by households who have children under the age

of 18. As expected, the estimated magnitude of the impact of child benefit on housing

related variables is very small. For example, the results from Table 5 suggest that a 100

Euro increase of child benefit increases the number of rooms in the homes of households

by about 0.04 to 0.07 rooms, depending on the number of children in the family. However,

on the margin additional child benefit results in improvements in the housing situation of

households with children. Moreover, there is a statistically significant labeling effect for child

benefit income: parents are more likely to use additional child benefit income than income

from other sources to improve their housing situation.

There are two margins along which the child benefit may impact the consumption of

housing services. First, a household may choose to buy instead of renting. Second, a house-

hold may “move up” to a larger or otherwise improved home. For those households that

19

currently rent their home, an increase in income could result in changes along either mar-

gin. The household may decide to pursue home ownership, or they may choose to rent a

larger residence. On the other hand, home owners are very unlikely to go back to renting

a home due to changes in child benefit income. Moreover, since selling a home is more

costly and time consuming than terminating a lease, it is also less likely that home owners

will purchase a larger home due to the change in child benefit income. Since there may

be systematic differences in how renters react to increased income compared to owners, I

estimate separate regressions for renters and owners. Additionally, I estimate the impact of

child benefit income on the amount of rent paid per square meter.21 Rent per square meter

can be interpreted as a measure of the quality of the home and an increase in child benefit

or household income may lead to households choosing to live in an identical size, but higher

quality home.

Tables 6 displays the results of estimating the impact of child benefit on rent per square

meter for households who rent their home. Tables 7-A/B and 8-A/B present the results of

estimating the impact of child benefit on the home size as well as the number of rooms in

the home separately for those households who rent their home and for households who own

their home. There is no statistically significant impact of child benefit on rent per square

meter paid by households. Although insignificant, the point estimate of the coefficient of

child benefit is positive in all specifications. With respect to the size of homes, comparing

the impact of child benefit income between renters and owners shows that the effects of child

benefit on apartment size are driven by renters. Table 7-A shows that a one Euro increase

in child benefit results in an increase in home size between 0.0093 and 0.0118 square meters

for renters, depending on the number of children in the household. Again, this effect is

driven by renters who have young children in the household. At the same time, there is no

statistically significant effect of child benefit on the size of the home for home owners (Table

21Home ownership is not always associated with regularly recurring payments since a household may bepaying a mortgage, but they may also have inherited the home, or paid off the home already in the past.Therefore, I do not have an equivalent to monthly housing payments available for owners.

20

7-B). Results are very similar for the impact of child benefit on the number of rooms in

the home for renters and owners. Table 8-A shows that the number of rooms in the home

increases for renters, and that this effect is driven by households with young children. On the

other hand, the impact of child benefit on the number of rooms in the home is statistically

insignificant in every specification for home owners, as shown in Table 8-B.

Parents Smoking and Drinking

Parents may use child benefit to support their smoking habits if child benefit income is

fungible. Therefore, I report results on whether the child benefit has any impact on the

probability that parents smoke. As described in Section 3, I use a panel of individuals

instead of households for these estimations and I include individual level fixed effects. Using

an individual-level fixed effects specification eliminates the need to include other important

determinants of smoking such as age and education (Chaloupka and Warner, 2000; Cutler

and Lleras-Muney, 2010) in the estimations. Since the average parent in my sample is 41

years old, there is not enough meaningful variation in education of individuals over time

and any difference in education between parents will be captured by the individual fixed

effect.22 Child benefit is paid to the parents of a family and is not assigned to a particular

individual parent in the family. Therefore, despite the fact that estimations are performed

using individual-level data, child benefit enters the estimation equations at the household

level.23

Table 9 presents estimates of the impact of child benefit income on the probability of

smoking. There is no statistically significant impact of child benefit on smoking participation

and there is no significant effect of household income on smoking participation either. While

the impact of both child benefit and household income are statistically insignificant, it is

interesting to note that the sign of the point estimate of the impact of household income is

22Moreover, Gohlmann (2007) finds that smoking initiation among older individuals in Germany is notaffected by education.

23Both parents in each household are included in the estimations individually, unless the family is headedby a single parent.

21

always negative, while there is no clear pattern for the coefficient of child benefit. If anything,

this could indicate that smoking may be an inferior good for parents with children. Table 10

shows that child benefit does not impact smoking on the intensive margin either. I report

only the impact of child benefit on the number of cigarettes smoked by parents who have

one child, parents who have one or two children, and parents with any number of children

because sample sizes are small for families with two children and for families with two or

three children. Conditional on smoking, there is no change in the number of cigarettes

consumed by parents in response to a change in child benefit, regardless of the number of

children.

Table 11 present results on drinking behavior. The outcome is a dummy variable that is

equal to 1 if the parent indicated to regularly consume either beer, wine, liquor, or mixed

drinks and zero otherwise.24 Blow et al. (2012) found that a large proportion of unexpected

increases in child benefit income in the UK were spent on alcohol. While I do not have data

on expenditures on alcohol, I am unable to find a significant impact on drinking behavior in

response to changes in child benefit income. Note, however that respondents to the SOEP

were asked about their drinking behavior only twice: in 2006 and 2008. During this time

there was no change in nominal child benefit and the only identifying variation in child

benefit comes from the fact that inflation decreased real child benefit between 2006 and

2008. Due to these data limitations results from the drinking behavior regressions should be

taken with caution.

Parents’ social activities

Individuals in the SOEP are also asked about leisure and social activities. In particular,

individuals indicate how frequently they attend pop concerts or go to the movies, attend

24The questionnaire includes separate questions for beer, wine, liquor, or mixed drinks. Respondentsare asked to indicate whether they consume each of these types of beverages “Regularly”, “Sometimes”,“Rarely”, or “Never”. I consider parents to regularly drink alcohol and code the dummy variable equal to1 if they indicated to “Regularly” consume any one or more of the above types of beverages. Results arerobust to using different classification schemes.

22

cultural events such as classical music concerts, the opera, public lectures, or the theater,

and how frequently they make excursions or take short vacation trips. All household members

who are at least 17 years old answer these questions. However, in order to investigate changes

in consumption patterns of parents in response to changes in child benefit, I consider only

parents in my estimations. If there is a labeling effect for child benefit, then we would not

expect child benefit income to be used for parents’ social and leisure activities. In particular,

since the child benefit is meant to ensure the basic needs of children are met, a labeling effect

would not earmark child benefit income for goods such as movie tickets or tickets to the opera.

In order to estimate the impact of child benefit on participation in social activities such as

going to the movies or attending pop concerts, and participating in cultural events such

as classical music concerts, the opera, public lectures, or theater performances, I create a

dummy variable for each individual parent that takes a value of one if the parent indicated

to participate in such activities at least once a month, and zero otherwise. Similarly, I

create a dummy variable that takes a value of one if the parent indicated to make excursions

or go on short vacation trips at least monthly. The results presented here are robust to

alternative definitions of this measure, for example a dummy variable indicating whether

parents participate weekly rather than monthly.

Table 12 shows that parents do indeed use child benefit income to attend pop concerts or

to go to the movies. The impact of child benefit on the probability of going to the movies or

to pop concerts at least once a month is positive and highly statistically significant. Based on

Table 12, a 10 Euro increase in child benefit results in an increased probability of attending

pop concerts or the movies at least once a month by 1.3–3 percentage points, depending on

the number of children in the household. The effect of child benefit is larger for parents who

have children that are all older than 18 years of age. The probability of going to the movies

or pop music concerts increases by 1.3–2.7 percentage points for an additional 10 Euros of

child benefit for households with young children. For households with older children, the

impact of child benefit is about four times larger. Depending of the number of children in the

23

household, the probability of going to the movies or attending pop concerts at least monthly

increases by 5.1–12 percentage points for a 10 Euro increase in child benefit. This means

that parents use child benefit for their own entertainment to a larger extent when children

are older, compared to when children are younger. Furthermore, the impact of an additional

Euro of child benefit on the probability that parents go to the movies or to pop concerts at

least once a month is statistically significantly larger than the impact of an additional Euro

of household income.

Similar results hold for parents participating in cultural events, such as going to classical

concerts, the opera, public lectures, or the theater. As shown in Table 13, parents are more

likely to attend such cultural events as child benefit increases. Regardless of the number of

children in the family, child benefit is used to a greater extent for opera, theater, or public

lectures by parents who have older children. The impact of child benefit on the probability

of participating in cultural events is similar in magnitude, yet somewhat smaller, compared

to the magnitude of the impact of child benefit on the probability of participating in other

social events such as the movies or pop music concerts.

Increases in child benefit are also associated with an increased probability of going on

excursions or short vacation trips at least once a month (Table 14). However, the estimated

impact of child benefit is relatively small compared to the estimates of other activities dis-

cussed above. Moreover, contrary to the impact of child benefit on the probability of going to

the movies or cultural events, the increases in the probability of going on excursions or short

vacation trips due to an increase in child benefit are driven by households with younger chil-

dren. Since excursions of parents with young children in the household include trips such as

visits to the zoo, museums, or amusement parks together with children, this result indicates

that child benefit is used for the benefit of children.

Parents respond individually to questions about how frequently they participate in the

activities discussed above. It is likely, though not necessary, that parents go to the movies,

concerts, and trips together as couples. In about 90 percent of cases, parents in the same

24

households agree on whether they engage in these activities at least monthly.25 The results

in Tables 12, 13, and 14 are robust to restricting the estimation sample to only those cases

in which parents both agree.26

5 Robustness Checks

In order to test the robustness of the results obtained in this paper, I estimate several alterna-

tive specifications. My main results are based on specifications that include the lagged value

of household income as a measure of other income. This avoids the potential bias introduced

into the model if child benefit income affects labor supply and therefore wage income in the

current period. Since the previous literature did not consider this potential confounding and

estimated specifications that included contemporaneous child benefit income and household

income, I examine whether any of my results change when I also use contemporaneous child

benefit income and household income such that

Outcomejtk = αkyjt + βkmjt + Zjtδk + µjk + νsk + τstk + εjtk, (5)

where outcomes are determined by child benefit (y) and current income (m).

Moreover, if additional child benefit income indeed causes a decrease in labor supply and

therefore a decrease in household income, then γ is negative in

mjt = γyjt + Zjtλ+ νjt, (6)

where mjt is household income (minus child benefit) in household j during year t, yjt is child

benefit in household j during year t, and Zjt is a vector of exogenous household character-

25Parents agree about going to the movies or pop music concerts at least monthly in 87% of all cases,agree about participating in cultural events at least monthly in 93% of all cases, and agree about makingexcursions or short trips at least monthly in 84% of all cases.

26Tables with results of restricting the estimation sample those cases in which parents both agree areomitted in the interest of space. Those results are available upon request.

25

istics. Substituting (6) back into Equation (5) results in

Outcomejtk = (αk + βkγ) yjt + Zjtδk + µjk + νsk + τstk + εjtk. (7)

Therefore, I also estimate reduced form specifications according to Equation (7) that

only include child benefit on the right hand side. Since γ is expected to be negative if there

is a significant impact of child benefit on household income, then for normal goods I expect

the reduced form coefficient to be smaller in magnitude compared to the estimates of the

main results. For inferior goods (when βk is negative) the magnitude of the reduced form

coefficient would be larger. In the interest of space, I only report results of the robustness

checks for regressions that include children of any age. Separate results for families with

young children and families with older children are available upon request.

Regarding per capita food expenditures, the results obtained by estimating Equation (5),

where I include contemporaneous household income rather than lagged household income,

provides very similar estimates compared to the main results (See Appendix Table 1). Fur-

thermore, the small differences between the estimates of the coefficient of child benefit do

not appear to be systematic. For households with only one child and households with two

or three children, the coefficient of child benefit is smaller when estimating Equation (5)

(reported in Appendix Table 1) in comparison to Equation (3) (Table 2), whereas for all

other household sizes the coefficient is larger. However, there is a systematic difference in

the magnitude of the household income variable. Using contemporaneous household income

and child benefit in the regression increases the point estimate of the impact of household

income on per capita food expenditures. The magnitude of the difference is not enough

to erase the labeling effect: An F-test of whether the coefficient of child benefit and the

coefficient of household income are equal rejects the null hypothesis with a p-value of < 0.00

regardless of family size.

The results of estimating the reduced form are also presented in Appendix Table 1. The

26

estimate of the impact of child benefit on per capita food expenditures from the reduced

form equation is very similar to the estimate from the main results. Interestingly, the impact

of child benefit on per capita food expenditures is larger in the reduced form specification

compared to the specifications that include a measure of household income. However, the

difference between the coefficients is small.

I also check whether my results are robust to running specifications that use the logarithm

of per capita food expenditures as the outcome variable. Appendix Table 2 shows that per

capita food expenditures increase between 0.08% and 0.25% for an additional Euro of child

benefit, depending on the number of children in the household. The impact of child benefit

is statistically significant for households with young children and households with children

of any age. Similar to the main results where I use the level of per capita food expenditure

as the outcome variable presented in Table 2, the impact of child benefit for households with

older children is statistically insignificant. The magnitude of the impact of child benefit is

also very similar in the log-linear specification compared to the main results. For example,

the estimated 0.08% increase in per capita food expenditures for an additional Euro in child

benefit for households with two children (Column 2 in Appendix Table 2) translates into an

increase in per capita food expenditures by 11.2 cents over a mean of 140.03 Euros. This is

very similar to the estimate of 12.99 cents per additional Euro in child benefit from Column

2 of Table 2.

Housing related results are also robust to using different measures of household income.

Appendix Tables 3 – 5 show the results of estimating the impact of child benefit on the

probability of being a renter, the size of the home in square meters, as well as the num-

ber of rooms in the home when using contemporaneous household income as the measure

of household income as depicted in Equation (5), and when estimating the reduced form

specification shown in Equation (7). The impact of child benefit on each of the outcomes

is consistent with the main results and numerically very similar. Appendix Table 5 shows

that the reduced form estimate of the impact of child benefit on the number of rooms in the

27

house is smaller than the estimate of the impact of child benefit when household income is

included in the regression. This is as expected, although the difference is very small.

Appendix Table 6 and Appendix Table 7 show that there is no impact of child benefit

on the probability that parents smoke and no impact of child benefit on the probability

that parents frequently consume alcohol regardless of the measure of household income.

With the exception of households with only one child present, the impact of household

income on smoking remains statistically insignificant when contemporaneous child benefit

and household income are used in the estimation. In the reduced form specifications, the

impact of child benefit on smoking is statistically insignificant regardless of the number of

children in the household, and there is no pattern in the sign of the point estimates.

Regarding parents’ social activities, Appendix Table 8 and Appendix Table 9 show that

there is little difference in the impact of child benefit between the main results based on Equa-

tion (3) and the results of estimating specifications including contemporaneous child benefit

and household income according to Equation (5). Estimating the reduced form specification

also produces results that are consistent with the main results. When using contempora-

neous child benefit and household income in the specification, the impact of an additional

Euro of household income on the probability of participating in cultural events is generally

positive and statistically significant. The impact of an additional Euro of household income

on the probability that parents go the movies or pop music concerts at least once a month

is also positive and statistically significant for households with two or three children. This

result is expected. However, in the specifications using the lagged value of household income,

the impact of lagged household income was statistically indistinguishable from zero in all

cases. Since going to the movie theater and participating in cultural events are expected to

be normal goods, the former result is more intuitive and the specification using contempo-

raneous child benefit and household income may be preferred to the specification that uses

the lagged value of household income.

Since the reduced form estimates, as well as the estimates obtained from regressions that

28

include contemporaneous child benefit and household income are qualitatively and quanti-

tatively similar to the main results, I conclude that the impact of child benefit on household

income through the channel of labor supply is negligible.

6 Conclusion

The German child benefit is intended to ensure that children’s needs in terms of nutrition

and housing (among other things) are met. Since the benefit is paid in cash and is fungible, I

investigate whether parents really do use the child benefit for its intended purpose. I find that

households primarily increase their per capita food expenditures in response to exogenous

increases in child benefit, which should improve nutrition. Households spend on average

between 49 and 74 cents out of every additional Euro of child benefit on food. Households

also use child benefit to improve their housing conditions: they are more likely to own their

home instead of renting, and are more likely to live in a larger home if they are renters.

I also find that there is a significant labeling effect for the child benefit, indicating that

households treat child benefit income differently compared to income from other sources.

An increase in child benefit income leads to larger improvements in nutrition and housing

conditions compared to an identical increase in general household income. This is consistent

with the evidence obtained in the previous literature for the Dutch child benefit. I find no

evidence that child benefit increases smoking of parents and, contrary to evidence from the

United Kingdom, I do not find that child benefit income causes parents to consume more

alcohol.

I find that parents use child benefit for their personal entertainment and social activities

that are unrelated to the well-being of children. As child benefit increases, parents are more

likely to go to the movies or to pop music concerts at least once a month, and they are more

likely to attend cultural events such as classical music concerts, the opera, public lectures,

or theater performances. Parents also use child benefit to make excursions and to take short

29

vacation trips.

Since parents may be eligible for child benefit up to the point when their child is 25

years old, I investigate whether parents of older children respond differently to changes in

child benefit. Per capita household expenditure on food does not increase for households

in which all children are older than 18 years, while it does for households with younger

children. Similarly, the impact of child benefit on housing related variables is not significant

for households with older children.

Interestingly, compared to the parents of younger children, parents whose children are

older than 18 years of age are more likely to use child benefit for their personal entertainment

activities such as going to the movies or pop music concerts, and they are more likely to

use child benefit to attend cultural events such as classical music concerts, the opera, public

lectures, or the theater. On the other hand, parents are more likely to use child benefit to

go on day trips or short vacations when children are young. Given that such trips include

outings to the zoo and amusement parks, this is not necessarily a detrimental outcome for

the well-being of young children.

Since it is the explicit intention of the policy maker to ensure the wellbeing of children in

terms of nutrition and housing, these results show that the child benefit does indeed have the

desired effect for families with non-adult children. Parents improve nutrition and housing

conditions in response to an increase in child benefit when children are young, and parents do

not use the child benefit to increase smoking or drinking. Once children are older, however,

parents no longer improve nutrition and housing conditions when child benefit increases.

Rather they use the child benefit for their own social activities and entertainment, i.e. they

increase expenditures on goods that are assignable directly to the parents, not the children.

30

Figure 1: Child Benefit Amounts, 1998-2009

31

Table 1: Summary Statistics

Panel (a) Variables Observed at the Household Level

Variable Description Mean Std. Dev Obs

Child Benefit Real Child Benefit in 2005e/month 261.79 134.54 39,598HH Income Real Net Houshold Income in

2005e/month2655.23 1424.60 38,084

Number of Children Number of Children in Household forwhich child benefit is received

1.7322 0.8377 39,598

Only Child Household has one child 0.4549 0.4980 39,598Two Children Household has two children 0.4012 0.4902 39,598Three Children Household has three children 0.1125 0.3160 39,598Four or more Children Household has at least four children 0.0313 0.1742 39,598Single Parent Single Parent 0.1387 0.3457 39598Age of Youngest Child Age of the youngest child in the house-

hold10.276 6.739 39,598

Food Expenditure Total per capita monthly expenditures onfood in the household

148.59 67.77 20,590

Renter Household rents or owns their residence(Rent=1/Own=0)

0.4543 0.4979 37,135

Size (m2) Size of the residence measured in squaremeters

112.74 41.98 37,136

Number of Rooms The number of rooms in the home largerthan 6 square meters

4.5681 1.6197 37,136

Rent/m2 Rent per square meter in real2005e/month

5.8539 1.9343 16,117

Unemployment Rate State-level unemployment rate 10.260 4.379 39,598Recession Country is in recession in current year 0.4783 0.4995 39,598

Panel (b) Variables Observed at the Individual Level

Variable Description Mean Std. Dev Obs

Age Age of Parent 40.88 8.33 70,875Smoking Parent Smokes 0.3535 0.4781 35,693Number of Cigarettes Number of Cigarettes smoked per day

among smokers16.2879 8.8358 7,985

Drinking Parent regularly consumes alcohol 0.1506 0.3577 11,159Movies or Pop Concerts Parents go to the movies or pop music con-

certs at least once a month0.2277 0.4194 44,089

Cultural Events Parents attend cultural events such as clas-sical music concerts, the opera, public lec-tures, or the theater at least once a month

0.1459 0.3531 44,111

Short Trips Parents make excursions or take short va-cation trips at least once a month

0.2577 0.4374 16,269

Note: Net HH Income does not include the Child Benefit. Real Child Benefit and real net household incomewere calculated using the overall consumer price index. Rent per square meter is deflated using the specific rent-and-utility price index. Summary statistics for housing related variables exclude households that have receivingsubsidized public housing.

32

Table 2: The Impact of Child Benefit on Per Capita Monthly Food Expenditures inHouseholds.

Per Person Real Monthly Household Food Expenditures

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Households with children of any age

Child Benefit 0.3971∗∗∗ 0.1299∗∗∗ 0.1482∗∗∗ 0.1614∗∗∗ 0.1829∗∗∗ 0.2070∗∗∗

(0.1091) (0.0391) (0.0358) (0.0323) (0.0303) (0.0272)

HH Income 0.0069∗∗∗ 0.0034∗ 0.0051∗∗∗ 0.0030∗ 0.0047∗∗∗ 0.0046∗∗∗

(0.0019) (0.0014) (0.0011) (0.0012) (0.0010) (0.0009)

p(CB = HH Income) 0.0004 0.0013 0.0001 0.0000 0.0000 0.0000Observations 6,942 6,856 13,798 8,803 15,745 16,295

Households with young children

Child Benefit 0.4304∗∗∗ 0.1376∗∗∗ 0.1587∗∗∗ 0.1666∗∗∗ 0.1879∗∗∗ 0.1984∗∗∗

(0.1305) (0.0405) (0.0376) (0.0332) (0.0315) (0.0281)

HH Income 0.0062∗ 0.0048∗∗∗ 0.0051∗∗∗ 0.0038∗∗ 0.0045∗∗∗ 0.0045∗∗∗

(0.0024) (0.0013) (0.0011) (0.0012) (0.0010) (0.0010)

p(CB = HH Income) 0.0012 0.0011 0.0000 0.0000 0.0000 0.0000Observations 5,067 6,292 11,359 8,182 13,249 13,793

Households with old children

Child Benefit 0.5283 0.3169 0.3071 0.1256 0.3028 0.2763(0.3718) (0.3532) (0.2172) (0.3003) (0.2043) (0.1998)

HH Income 0.0005 -0.0021 0.0010 -0.0022 0.0007 0.0006(0.0043) (0.0027) (0.0027) (0.0027) (0.0027) (0.0026)

p(CB = HH Income) 0.1565 0.3680 0.1588 0.6710 0.1394 0.1679Observations 1,406 355 1,761 391 1,797 1,800

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05, ∗∗∗p < 0.01.p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coefficient of ChildBenefit (CB) equals the coefficient of other household income. Households with young children have atleast one child under the age of 18 living in the households. Families with older children have childrenwho are all at least 18 years of age. Household income does not include the Child Benefit. Both ChildBenefit and Household Income are Measured in Real 2005 e. Household income is lagged by one year.Other controls included in each specification are single parent indicators, a quadratic term in the age ofthe youngest child, the state level unemployment rate, an dummy variables for whether the country wasin a recession during a sample year, dummy variables for states as well as state specific year trends. Spec-ifications for Columns (3)-(6) also control for the number of children in the household.

33

Table 3: The Impact of Child Benefit on the Probability that Households Rent TheirHome.

Probability of Being a Renter

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Households with children of any age

Child Benefit -0.0001∗∗ -0.0001 -0.0001∗∗∗ -0.0001∗ -0.0001∗∗∗ -0.0001∗∗∗

(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

HH Income -0.0000∗∗∗ -0.0000∗∗ -0.0000∗∗∗ -0.0000∗∗ -0.0000∗∗∗ -0.0000∗∗∗

(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.0252 0.1539 0.0007 0.0522 0.0001 0.0002Observations 10,829 11,628 22,457 14,875 25,704 26,581

Households with young children

Child Benefit -0.0002∗ -0.0001 -0.0001∗∗ -0.0001 -0.0001∗∗ -0.0001∗∗

(0.0001) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

HH Income -0.0000∗∗ -0.0000∗∗ -0.0000∗∗∗ -0.0000∗∗ -0.0000∗∗∗ -0.0000∗∗∗

(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.0394 0.3081 0.0169 0.1197 0.0064 0.0095Observations 7,557 10,595 18,152 13,737 21,294 22,161

Households with old children

Child Benefit -0.0000 -0.0000 0.0000 -0.0000 0.0000 0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

HH Income -0.0000∗ -0.0000 -0.0000 -0.0000 -0.0000 -0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.8705 0.9440 0.5112 0.9891 0.4395 0.5022Observations 1,775 425 2,200 463 2,238 2,242

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05, ∗∗∗p < 0.01.p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coefficient of Child Ben-efit (CB) equals the coefficient of other household income. Households with young children have at leastone child under the age of 18 living in the households. Families with older children have children who areall at least 18 years of age. Household income does not include the Child Benefit. Both Child Benefit andHousehold Income are Measured in Real 2005 e. Child Benefit is lagged by one year. Household income islagged by two years. Other controls included in each specification are single parent indicators, a quadraticterm in the age of the youngest child, the state level unemployment rate, an dummy variables for whetherthe country was in a recession during a sample year, dummy variables for states as well as state specificyear trends.Specifications for Columns (3)-(6) also control for the number of children in the household.

34

Table 4: The Impact of Child Benefit on the Size (Square Meters) of Homes of House-holds.

Size of Home in Square Meters

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Households with children of any age

Child Benefit 0.0095∗ 0.0079 0.0133∗∗∗ 0.0094∗ 0.0145∗∗∗ 0.0128∗∗∗

(0.0046) (0.0045) (0.0037) (0.0043) (0.0036) (0.0034)

HH Income 0.0008∗ 0.0006 0.0009∗∗ 0.0009∗ 0.0011∗∗∗ 0.0012∗∗∗

(0.0003) (0.0005) (0.0003) (0.0004) (0.0003) (0.0003)

p(CB = HH Income) 0.0586 0.1045 0.0008 0.0516 0.0002 0.0007Observations 10,829 11,628 22,457 14,875 25,704 26,581

Households with young children

Child Benefit 0.0078 0.0089 0.0131∗∗ 0.0093 0.0127∗∗ 0.0108∗∗

(0.0056) (0.0051) (0.0044) (0.0048) (0.0040) (0.0039)

HH Income 0.0006 0.0005 0.0008∗ 0.0010∗ 0.0011∗∗ 0.0011∗∗∗

(0.0004) (0.0006) (0.0004) (0.0005) (0.0003) (0.0003)

p(CB = HH Income) 0.1960 0.0949 0.0049 0.0861 0.0041 0.0145Observations 7,557 10,595 18,152 13,737 21,294 22,161

Households with old children

Child Benefit 0.0036 -0.0033 -0.0009 0.0140 0.0035 0.0037(0.0086) (0.0124) (0.0065) (0.0125) (0.0064) (0.0064)

HH Income -0.0000 0.0001 0.0001 0.0003 0.0002 0.0002(0.0006) (0.0004) (0.0003) (0.0004) (0.0003) (0.0003)

p(CB = HH Income) 0.6754 0.7775 0.8750 0.2706 0.6073 0.5840Observations 1,775 425 2,200 463 2,238 2,242

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05,∗∗∗p < 0.01. p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coef-ficient of Child Benefit (CB) equals the coefficient of other household income. Households with youngchildren have at least one child under the age of 18 living in the households. Families with older childrenhave children who are all at least 18 years of age. Household income does not include the Child Benefit.Both Child Benefit and Household Income are Measured in Real 2005 e. Child Benefit is lagged by oneyear. Household income is lagged by two years. Other controls included in each specification are singleparent indicators, a quadratic term in the age of the youngest child, the state level unemployment rate,an dummy variables for whether the country was in a recession during a sample year, dummy variablesfor states as well as state specific year trends. Specifications for Columns (3)-(6) also control for thenumber of children in the household.

35

Table 5: The Impact of Child Benefit on the Number of Rooms in the Homes of House-holds.

Number of Rooms

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Households with children of any age

Child Benefit 0.0003 0.0003∗ 0.0006∗∗∗ 0.0004∗∗ 0.0007∗∗∗ 0.0006∗∗∗

(0.0002) (0.0002) (0.0001) (0.0002) (0.0001) (0.0001)

HH Income 0.0000 0.0000∗∗ 0.0000∗∗∗ 0.0001∗∗∗ 0.0001∗∗∗ 0.0001∗∗∗

(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.1662 0.0555 0.0002 0.0145 0.0000 0.0000Observations 10,829 11,628 22,457 14,875 25,704 26,581

Households with young children

Child Benefit 0.0003 0.0004∗ 0.0005∗∗ 0.0004∗ 0.0006∗∗∗ 0.0005∗∗∗

(0.0003) (0.0002) (0.0002) (0.0002) (0.0001) (0.0001)

HH Income 0.0000 0.0000 0.0000∗ 0.0000∗∗ 0.0000∗∗∗ 0.0000∗∗∗

(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.2278 0.0301 0.0039 0.0290 0.0006 0.0025Observations 7,557 10,595 18,152 13,737 21,294 22,161

Households with old children

Child Benefit 0.0003 -0.0004 0.0004 -0.0004 0.0004 0.0004(0.0004) (0.0007) (0.0003) (0.0005) (0.0003) (0.0003)

HH Income 0.0001∗ 0.0001∗∗∗ 0.0001∗∗ 0.0001∗∗∗ 0.0001∗∗ 0.0001∗∗

(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.6018 0.4553 0.2173 0.3300 0.2782 0.2595Observations 1,775 425 2,200 463 2,238 2,242

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05,∗∗∗p < 0.01. p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coef-ficient of Child Benefit (CB) equals the coefficient of other household income. Households with youngchildren have at least one child under the age of 18 living in the households. Families with older childrenhave children who are all at least 18 years of age. Household income does not include the Child Benefit.Both Child Benefit and Household Income are Measured in Real 2005 e. Child Benefit is lagged by oneyear. Household income is lagged by two years. Other controls included in each specification are singleparent indicators, a quadratic term in the age of the youngest child, the state level unemployment rate,an dummy variables for whether the country was in a recession during a sample year, dummy variablesfor states as well as state specific year trends. Specifications for Columns (3)-(6) also control for thenumber of children in the household.

36

Table 6: The Impact of Child Benefit on the Amount of Rent Paid per Square Meter.

Rent Per Square Meter

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Households with children of any age

Child Benefit 0.0002 0.0002 0.0002 0.0004 0.0003 0.0002(0.0004) (0.0003) (0.0002) (0.0003) (0.0002) (0.0002)

HH Income 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.7124 0.5425 0.5487 0.1294 0.2816 0.4419Observations 5,285 4,281 9,566 5,498 10,783 11,129

Households with young children

Child Benefit 0.0004 0.0002 0.0002 0.0005 0.0004 0.0003(0.0004) (0.0003) (0.0002) (0.0003) (0.0002) (0.0002)

HH Income 0.0001 -0.0000 0.0000 -0.0000 0.0000 0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.4535 0.4669 0.4916 0.0972 0.1228 0.1598Observations 3,940 3,974 7,914 5,170 9,110 9,455

Households with old children

Child Benefit 0.0022 0.0009 0.0016 -0.0004 0.0014 0.0014(0.0012) (0.0023) (0.0009) (0.0020) (0.0008) (0.0008)

HH Income -0.0000 0.0004∗∗ 0.0001 0.0003∗∗ 0.0000 0.0000(0.0001) (0.0001) (0.0000) (0.0001) (0.0000) (0.0000)

p(CB = HH Income) 0.0692 0.8277 0.0729 0.7257 0.0902 0.0902Observations 684 116 800 123 807 807

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05,∗∗∗p < 0.01. p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coef-ficient of Child Benefit (CB) equals the coefficient of other household income. Households with youngchildren have at least one child under the age of 18 living in the households. Families with older chil-dren have children who are all at least 18 years of age. Household income does not include the ChildBenefit. Both Child Benefit and Household Income are Measured in Real 2005 e. Child Benefit islagged by one year. Household income is lagged by two years. Other controls included in each speci-fication are single parent indicators, a quadratic term in the age of the youngest child, the state levelunemployment rate, an dummy variables for whether the country was in a recession during a sampleyear, dummy variables for states as well as state specific year trends. Specifications for Columns (3)-(6)also control for the number of children in the household.

37

Table 7-A: The Impact of Child Benefit on the Size (Square Meters) of Homes for House-holds that Rent their Home.

Size of Home in Square Meters (Renters)

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Households with children of any age

Child Benefit 0.0058 0.0076 0.0093∗ 0.0090∗ 0.0118∗∗ 0.0117∗∗∗

(0.0051) (0.0049) (0.0037) (0.0042) (0.0036) (0.0030)

HH Income 0.0007 0.0013∗∗ 0.0012∗∗∗ 0.0015∗∗∗ 0.0014∗∗∗ 0.0014∗∗∗

(0.0004) (0.0005) (0.0003) (0.0004) (0.0003) (0.0003)

p(CB = HH Income) 0.3198 0.2014 0.0257 0.0778 0.0037 0.0007Observations 5,488 4,522 10,010 5,787 11,275 11,626

Households with young children

Child Benefit 0.0029 0.0061 0.0081∗ 0.0082 0.0106∗∗ 0.0104∗∗

(0.0053) (0.0051) (0.0040) (0.0046) (0.0040) (0.0032)

HH Income 0.0006 0.0016∗∗ 0.0012∗∗ 0.0018∗∗∗ 0.0014∗∗∗ 0.0015∗∗∗

(0.0005) (0.0005) (0.0004) (0.0004) (0.0003) (0.0003)

p(CB = HH Income) 0.6656 0.3684 0.0899 0.1598 0.0216 0.0065Observations 4,105 4,203 8,308 5,447 9,552 9,902

Households with old children

Child Benefit -0.0092 -0.0018 0.0002 0.0007 0.0019 0.0019(0.0091) (0.0034) (0.0064) (0.0039) (0.0059) (0.0059)

HH Income -0.0001 -0.0009∗∗ -0.0003 -0.0010∗∗∗ -0.0003 -0.0003(0.0005) (0.0002) (0.0004) (0.0002) (0.0004) (0.0004)

p(CB = HH Income) 0.3221 0.7906 0.9452 0.6775 0.7098 0.7098Observations 704 121 825 128 832 832

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05,∗∗∗p < 0.01. p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coef-ficient of Child Benefit (CB) equals the coefficient of other household income. Households with youngchildren have at least one child under the age of 18 living in the households. Families with older childrenhave children who are all at least 18 years of age. Household income does not include the Child Benefit.Both Child Benefit and Household Income are Measured in Real 2005 e. Child Benefit is lagged by oneyear. Household income is lagged by two years. Other controls included in each specification are singleparent indicators, a quadratic term in the age of the youngest child, the state level unemployment rate,an dummy variables for whether the country was in a recession during a sample year, dummy variables forstates as well as state specific year trends. Specifications for Columns (3)-(6) also control for the numberof children in the household.

38

Table 7-B: The Impact of Child Benefit on the Size (Square Meters) of Homes forHouseholds that Own their Home.

Size of Home in Square Meters (Owners)

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Households with children of any age

Child Benefit 0.0022 -0.0022 0.0043 0.0012 0.0068 0.0058(0.0071) (0.0052) (0.0053) (0.0049) (0.0051) (0.0051)

HH Income -0.0003 -0.0004 -0.0002 -0.0002 -0.0001 -0.0000(0.0003) (0.0005) (0.0003) (0.0004) (0.0003) (0.0003)

p(CB = HH Income) 0.7243 0.7226 0.3897 0.7898 0.1761 0.2515Observations 5,341 7,106 12,447 9,088 14,429 14,955

Households with young children

Child Benefit -0.0029 0.0007 0.0037 0.0014 0.0049 0.0040(0.0100) (0.0059) (0.0063) (0.0055) (0.0057) (0.0060)

HH Income -0.0006 -0.0008 -0.0008 -0.0004 -0.0004 -0.0004(0.0005) (0.0006) (0.0004) (0.0005) (0.0004) (0.0004)

p(CB = HH Income) 0.8197 0.7911 0.4705 0.7492 0.3558 0.4665Observations 3,452 6,392 9,844 8,290 11,742 12,259

Households with old children

Child Benefit -0.0011 -0.0084 -0.0043 0.0166 0.0018 0.0021(0.0106) (0.0148) (0.0084) (0.0161) (0.0086) (0.0086)

HH Income -0.0002 0.0001 0.0001 0.0005 0.0002 0.0002(0.0007) (0.0005) (0.0003) (0.0005) (0.0003) (0.0003)

p(CB = HH Income) 0.9348 0.5609 0.5999 0.3149 0.8474 0.8257Observations 1,071 304 1,375 335 1,406 1,410

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05,∗∗∗p < 0.01. p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coef-ficient of Child Benefit (CB) equals the coefficient of other household income. Households with youngchildren have at least one child under the age of 18 living in the households. Families with older chil-dren have children who are all at least 18 years of age. Household income does not include the ChildBenefit. Both Child Benefit and Household Income are Measured in Real 2005 e. Child Benefit islagged by one year. Household income is lagged by two years. Other controls included in each speci-fication are single parent indicators, a quadratic term in the age of the youngest child, the state levelunemployment rate, an dummy variables for whether the country was in a recession during a sampleyear, dummy variables for states as well as state specific year trends. Specifications for Columns (3)-(6)also control for the number of children in the household.

39

Table 8-A: The Impact of Child Benefit on the Number of Rooms in Homes of House-holds that Rent their Home.

Number of Rooms (Renters)

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Households with children of any age

Child Benefit 0.0002 0.0004 0.0004∗∗ 0.0003 0.0005∗∗ 0.0005∗∗

(0.0002) (0.0002) (0.0002) (0.0002) (0.0001) (0.0001)

HH Income 0.0000 0.0001 0.0000∗ 0.0000 0.0000∗ 0.0000∗∗

(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.3403 0.1204 0.0077 0.1659 0.0027 0.0028Observations 5,488 4,522 10,010 5,787 11,275 11,626

Households with young children

Child Benefit 0.0002 0.0003 0.0004∗ 0.0002 0.0004∗ 0.0004∗

(0.0002) (0.0002) (0.0002) (0.0002) (0.0002) (0.0002)

HH Income 0.0000 0.0001 0.0000∗ 0.0000 0.0000∗ 0.0000∗

(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.4839 0.2508 0.0358 0.3961 0.0247 0.0326Observations 4,105 4,203 8,308 5,447 9,552 9,902

Households with old children

Child Benefit -0.0002 0.0017 0.0002 0.0013 0.0002 0.0002(0.0006) (0.0010) (0.0004) (0.0008) (0.0003) (0.0003)

HH Income 0.0000 0.0000 0.0000∗ 0.0000 0.0000∗ 0.0000∗

(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.6309 0.0983 0.6319 0.1038 0.6612 0.6612Observations 704 121 825 128 832 832

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05,∗∗∗p < 0.01. p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coef-ficient of Child Benefit (CB) equals the coefficient of other household income. Households with youngchildren have at least one child under the age of 18 living in the households. Families with older chil-dren have children who are all at least 18 years of age. Household income does not include the ChildBenefit. Both Child Benefit and Household Income are Measured in Real 2005 e. Child Benefit islagged by one year. Household income is lagged by two years. Other controls included in each speci-fication are single parent indicators, a quadratic term in the age of the youngest child, the state levelunemployment rate, an dummy variables for whether the country was in a recession during a sampleyear, dummy variables for states as well as state specific year trends. Specifications for Columns (3)-(6)also control for the number of children in the household.

40

Table 8-B: The Impact of Child Benefit on the Number of Rooms in Homes of House-holds that Own their Home.

Number of Rooms (Owners)

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Households with children of any age

Child Benefit -0.0001 0.0002 0.0003 0.0004 0.0005∗∗ 0.0005∗

(0.0004) (0.0002) (0.0002) (0.0002) (0.0002) (0.0002)

HH Income -0.0000 0.0000 0.0000 0.0000 0.0000 0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.7191 0.5436 0.1090 0.1134 0.0109 0.0154Observations 5,341 7,106 12,447 9,088 14,429 14,955

Households with young children

Child Benefit -0.0003 0.0003 0.0002 0.0004 0.0004 0.0003(0.0006) (0.0002) (0.0002) (0.0002) (0.0002) (0.0002)

HH Income -0.0000 0.0000 -0.0000 0.0000 -0.0000 -0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.6655 0.1955 0.4259 0.0946 0.0689 0.0995Observations 3,452 6,392 9,844 8,290 11,742 12,259

Households with old children

Child Benefit -0.0001 -0.0010 0.0001 -0.0008 0.0001 0.0001(0.0006) (0.0008) (0.0004) (0.0006) (0.0004) (0.0004)

HH Income 0.0000 0.0001∗∗∗ 0.0000∗ 0.0001∗∗∗ 0.0000∗ 0.0000∗

(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.8545 0.1519 0.8071 0.1725 0.9137 0.8823Observations 1,071 304 1375 335 1,406 1,410

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05,∗∗∗p < 0.01. p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coef-ficient of Child Benefit (CB) equals the coefficient of other household income. Households with youngchildren have at least one child under the age of 18 living in the households. Families with older childrenhave children who are all at least 18 years of age. Household income does not include the Child Benefit.Both Child Benefit and Household Income are Measured in Real 2005 e. Child Benefit is lagged by oneyear. Household income is lagged by two years. Other controls included in each specification are singleparent indicators, a quadratic term in the age of the youngest child, the state level unemployment rate,an dummy variables for whether the country was in a recession during a sample year, dummy variablesfor states as well as state specific year trends. Specifications for Columns (3)-(6) also control for thenumber of children in the household.

41

Table 9: The Impact of Child Benefit on the Probability that Parents Smoke.

Smoking Participation

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Households with children of any age

Child Benefit 0.0002 0.0001 0.0002 -0.0001 -0.0001 -0.0001(0.0006) (0.0002) (0.0002) (0.0002) (0.0002) (0.0001)

HH Income -0.0000 -0.0000 -0.0000 -0.0000 -0.0000 -0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.7057 0.5492 0.4000 0.4765 0.6204 0.7265Observations 12,756 13,157 25,913 16,973 29,729 30,773

Households with young children

Child Benefit 0.0002 0.0001 0.0001 -0.0002 -0.0002 -0.0001(0.0007) (0.0002) (0.0002) (0.0002) (0.0002) (0.0002)

HH Income 0.0000 -0.0000 -0.0000 -0.0000 -0.0000 -0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.7791 0.6457 0.7993 0.4135 0.2854 0.3948Observations 9,127 12,095 21,222 15,808 24,935 25,969

Households with old children

Child Benefit -0.0004 0.0018 0.0008 0.0005 0.0003 0.0006(0.0014) (0.0020) (0.0008) (0.0019) (0.0008) (0.0008)

HH Income -0.0000∗ 0.0000 -0.0000 -0.0000 -0.0000 -0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.7834 0.3769 0.3575 0.8085 0.6925 0.4418Observations 2,785 685 3,470 757 3,542 3,550

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05,∗∗∗p < 0.01. p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coef-ficient of Child Benefit (CB) equals the coefficient of other household income. Households with youngchildren have at least one child under the age of 18 living in the households. Families with older childrenhave children who are all at least 18 years of age. Household income does not include the Child Benefit.Both Child Benefit and Household Income are Measured in Real 2005 e. Household income is laggedby one year. Other controls included in each specification are single parent indicators, a quadratic termin the age of the youngest child, the state level unemployment rate, an dummy variables for whetherthe country was in a recession during a sample year, dummy variables for states as well as state specificyear trends. Specifications for Columns (3)-(6) also control for the number of children in the household.

42

Table 10: The Impact of Child Benefit on the Numberof Cigarettes that Parents Smoke.

Number of Cigarettes Per Day

(1) (2) (3)

Families withAny

1 1 or 2 Number ofChild Children Children

Households with children of any age

Child Benefit -0.4689 -0.0405 -0.0139(0.5989) (0.0291) (0.0144)

HH Income 0.0001 0.0000 0.0000(0.0002) (0.0001) (0.0001)

p(CB = HH Income) 0.4336 0.1638 0.3353Observations 3,185 5,861 6,907

Households with young children

Child Benefit -1.0667 -0.0535 -0.0194(0.6309) (0.0330) (0.0153)

HH Income 0.0002 0.0001 0.0001(0.0002) (0.0001) (0.0001)

p(CB = HH Income) 0.0911 0.1045 0.2020Observations 2,371 4,845 5,875

Households with old children

Child Benefit -2.0466 -0.0173 0.0634(1.3411) (0.1929) (0.1251)

HH Income -0.0003 -0.0001 -0.0001(0.0005) (0.0005) (0.0004)

p(CB = HH Income) 0.1277 0.9290 0.6121Observations 604 729 740

Note: Standard errors clustered at the household level are inparenthesis. ∗p < 0.1, ∗∗p < 0.05, ∗∗∗p < 0.01. p(CB = HH In-come) is the p-value of an F-test testing the null hypothesis thatthe coefficient of Child Benefit (CB) equals the coefficient of otherhousehold income. Households with young children have at leastone child under the age of 18 living in the households. Familieswith older children have children who are all at least 18 years ofage. Household income does not include the Child Benefit. BothChild Benefit and Household Income are Measured in Real 2005e. Other controls included in each specification are single parentindicators, a quadratic term in the age of the youngest child, thestate level unemployment rate, an dummy variables for whetherthe country was in a recession during a sample year, dummy vari-ables for states as well as state specific year trends. Householdincome is lagged by one year. Specifications for Columns (2) and(3) also control for the number of children in the household.

43

Table 11: The Impact of Child Benefit on the Probability that Parents Regularly DrinkAlcohol.

Drinking Alcohol Regularly

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Households with children of any age

Child Benefit -0.0357 0.0467 -0.0005 0.0042 0.0016 0.0007(0.0750) (0.0758) (0.0019) (0.0029) (0.0013) (0.0010)

HH Income 0.0000∗ -0.0000 0.0000 -0.0000 0.0000 0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.6339 0.5379 0.8082 0.1485 0.2289 0.4567Observations 3,942 3,967 7,909 5,074 9,016 9,295

Households with young children

Child Benefit -0.1051 0.0862 0.0010 0.0027 0.0021 0.0010(0.0789) (0.0749) (0.0021) (0.0029) (0.0015) (0.0010)

HH Income 0.0000∗ -0.0000∗ -0.0000 -0.0000∗ -0.0000 0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.1832 0.2496 0.6430 0.3507 0.1489 0.3537Observations 2,635 3,521 6,156 4,581 7,216 7,491

Households with old children

Child Benefit -0.0167 -0.0460 0.0045 0.0007 0.0029 0.0030(0.0337) (0.0376) (0.0101) (0.0010) (0.0082) (0.0043)

HH Income 0.0000 0.0001 0.0000 0.0001 0.0000 0.0000(0.0000) (0.0001) (0.0000) (0.0001) (0.0000) (0.0000)

p(CB = HH Income) 0.6201 0.2207 0.6595 0.5194 0.7225 0.4929Observations 1,031 310 1,341 339 1,370 1,372

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05,∗∗∗p < 0.01. p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coef-ficient of Child Benefit (CB) equals the coefficient of other household income. Households with youngchildren have at least one child under the age of 18 living in the households. Families with older childrenhave children who are all at least 18 years of age. Household income does not include the Child Benefit.Both Child Benefit and Household Income are Measured in Real 2005 e. Household income is laggedby one year. Other controls included in each specification are single parent indicators, a quadratic termin the age of the youngest child, the state level unemployment rate, an dummy variables for whetherthe country was in a recession during a sample year, dummy variables for states as well as state specificyear trends. Specifications for Columns (3)-(6) also control for the number of children in the household.

44

Table 12: The Impact of Child Benefit on the Probability that Parents Go to the Moviesor Attend Pop Music Concerts at least Once per Month

Going to the Movies or Pop Music Concerts Once a Month

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Households with children of any age

Child Benefit 0.0030∗∗∗ 0.0013∗∗∗ 0.0013∗∗∗ 0.0017∗∗∗ 0.0016∗∗∗ 0.0015∗∗∗

(0.0004) (0.0002) (0.0002) (0.0002) (0.0001) (0.0001)

HH Income -0.0000 0.0000 -0.0000 0.0000 -0.0000 -0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000Observations 15,952 16,394 32,346 21,120 37,072 38,364

Households with young children

Child Benefit 0.0027∗∗∗ 0.0013∗∗∗ 0.0013∗∗∗ 0.0017∗∗∗ 0.0016∗∗∗ 0.0016∗∗∗

(0.0006) (0.0002) (0.0002) (0.0002) (0.0002) (0.0001)

HH Income -0.0000 0.0000 0.0000 0.0000 0.0000 -0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000Observations 11,326 14,962 26,288 19,529 30,855 32,136

Households with old children

Child Benefit 0.0123∗∗∗ 0.0051∗∗ 0.0067∗∗∗ 0.0055∗∗∗ 0.0066∗∗∗ 0.0068∗∗∗

(0.0016) (0.0017) (0.0009) (0.0015) (0.0009) (0.0009)

HH Income -0.0000 0.0000 0.0000 0.0000 0.0000 0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.0000 0.0024 0.0000 0.0003 0.0000 0.0000Observations 3,546 953 4,499 1,053 4,599 4,605

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05, ∗∗∗p < 0.01.p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coefficient of ChildBenefit (CB) equals the coefficient of other household income. Households with young children have atleast one child under the age of 18 living in the households. Families with older children have childrenwho are all at least 18 years of age. Household income does not include the Child Benefit. Both ChildBenefit and Household Income are Measured in Real 2005 e. Household income is lagged by one year.Other controls included in each specification are single parent indicators, a quadratic term in the age ofthe youngest child, the state level unemployment rate, an dummy variables for whether the country wasin a recession during a sample year, dummy variables for states as well as state specific year trends. Spec-ifications for Columns (3)-(6) also control for the number of children in the household.

45

Table 13: The Impact of Child Benefit on Cultural Events Attended by Parents (Opera,Classical Music Concerts, Public Lectures, Theater.)

Attending Cultural Events (Opera, Classical Music Concerts,Public Lectures, Theater) Once a Month

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Households with children of any age

Child Benefit 0.0023∗∗∗ 0.0011∗∗∗ 0.0011∗∗∗ 0.0016∗∗∗ 0.0014∗∗∗ 0.0013∗∗∗

(0.0003) (0.0001) (0.0001) (0.0001) (0.0001) (0.0001)

HH Income -0.0000 -0.0000 -0.0000 -0.0000 -0.0000 -0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000Observations 15,963 16,396 32,359 21,122 37,085 38,379

Households with young children

Child Benefit 0.0025∗∗∗ 0.0009∗∗∗ 0.0010∗∗∗ 0.0015∗∗∗ 0.0014∗∗∗ 0.0013∗∗∗

(0.0004) (0.0002) (0.0001) (0.0001) (0.0001) (0.0001)

HH Income -0.0000∗ -0.0000 -0.0000 -0.0000 -0.0000 -0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000Observations 11,333 14,963 26,296 19,530 30,863 32,146

Households with old children

Child Benefit 0.0096∗∗∗ 0.0052∗∗ 0.0062∗∗∗ 0.0067∗∗∗ 0.0066∗∗∗ 0.0063∗∗∗

(0.0014) (0.0018) (0.0009) (0.0016) (0.0009) (0.0009)

HH Income -0.0000 0.0000 0.0000 0.0000 0.0000 0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.0000 0.0037 0.0000 0.0000 0.0000 0.0000Observations 3,551 953 4,504 1,053 4,604 4,610

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05, ∗∗∗p < 0.01.p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coefficient of ChildBenefit (CB) equals the coefficient of other household income. Households with young children have atleast one child under the age of 18 living in the households. Families with older children have childrenwho are all at least 18 years of age. Household income does not include the Child Benefit. Both ChildBenefit and Household Income are Measured in Real 2005 e. Household income is lagged by one year.Other controls included in each specification are single parent indicators, a quadratic term in the age ofthe youngest child, the state level unemployment rate, an dummy variables for whether the country wasin a recession during a sample year, dummy variables for states as well as state specific year trends. Spec-ifications for Columns (3)-(6) also control for the number of children in the household.

46

Table 14: The Impact of Child Benefit on the Probability that Parents Make Excur-sions or Go on Short Vacation Trips at least Once per Month.

Making Excursions or Going on Vacation Trips Once a Month

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Households with children of any age

Child Benefit 0.0012 0.0008 0.0010∗∗ 0.0007∗ 0.0008∗∗ 0.0006∗

(0.0009) (0.0004) (0.0003) (0.0003) (0.0003) (0.0003)

HH Income -0.0000 0.0000 0.0000 0.0000 0.0000 0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.2038 0.0780 0.0025 0.0331 0.0054 0.0208Observations 5,833 5,948 11,781 7,689 13,522 13,975

Households with young children

Child Benefit 0.0028∗ 0.0007 0.0011∗∗ 0.0006 0.0008∗ 0.0006∗

(0.0013) (0.0005) (0.0004) (0.0004) (0.0003) (0.0003)

HH Income -0.0000 0.0000 -0.0000 0.0000 0.0000 0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.0264 0.1566 0.0059 0.1349 0.0181 0.0352Observations 4,147 5,401 9,548 7,079 11,226 11,674

Households with old children

Child Benefit -0.0044 0.0068 -0.0008 0.0068 -0.0009 -0.0008(0.0035) (0.0048) (0.0018) (0.0048) (0.0018) (0.0017)

HH Income 0.0003∗∗ 0.0001 0.0002∗ 0.0001 0.0002∗ 0.0002∗

(0.0001) (0.0001) (0.0001) (0.0001) (0.0001) (0.0001)

p(CB = HH Income) 0.1912 0.1638 0.5994 0.1631 0.5540 0.5566Observations 1,277 346 1,623 384 1,661 1,663

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05,∗∗∗p < 0.01. p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coef-ficient of Child Benefit (CB) equals the coefficient of other household income. Households with youngchildren have at least one child under the age of 18 living in the households. Families with older childrenhave children who are all at least 18 years of age. Household income does not include the Child Benefit.Both Child Benefit and Household Income are Measured in Real 2005 e. Household income is laggedby one year. Other controls included in each specification are single parent indicators, a quadratic termin the age of the youngest child, the state level unemployment rate, an dummy variables for whetherthe country was in a recession during a sample year, dummy variables for states as well as state specificyear trends. Specifications for Columns (3)-(6) also control for the number of children in the household.

47

References

Aaronson, D. (2000). A note on the benefits of homeownership. Journal of Urban Economics,47(3):356–369.

Becker, G. S. (1981). A Treatise on the Family. Harvard University Press, Cambridge.

Blow, L., Walker, I., and Zhu, Y. (2012). Who benefits from child benefit. Economic Inquiry,50(1):153–170.

Blundell, R. and Hoynes, H. W. (2004). Has in-work benefit reform helped the labor market?In Card, D., Blundell, R., and Freeman, R. B., editors, Seeking a Premier Economy: TheEconomic Effects of British Economic Reforms, 1980-2000. University of Chicago Press,Chicago.

Browning, M., Bourguignon, F., Chiappori, P.-A., and Lechene, V. (1994). Incomes andoutcomes: A structural model of intrahousehold allocation. Journal of Political Economy,102(6):1067–1096.

Browning, M. and Meghir, C. (1991). The effects of male and female labor supply oncommodity demands. Econometrica, 59(4):925–951.

Bundesministerium der Finanzen (2010). Datensammlung zur Steuerpolitik, Ausgabe 2010.Technical report, Bundesministerium der Finanzen - Referat Offentlichkeitsarbeit, Berlin.

Bundesverfassungsgericht (1998). BVerfG, 2 BvR 1057/91.

Chaloupka, F. J. and Grossman, M. (1997). Price, tobacco control policies and youth smok-ing. Journal of Health Economics, 16(3):359–373.

Chaloupka, F. J. and Warner, K. E. (2000). Chapter 29: The economics of smoking. InCulyer, A. J. and Newhouse, J. P., editors, Handbook of Health Economics, volume 1-B,pages 1539–1627. Elsevier.

Cheng, K.-W. and Kenkel, D. S. (2010). U.S. cigarette demand: 1944-2004. The B.E.Journal of Economic Analysis & Policy, 10(1).

Cutler, D. M. and Lleras-Muney, A. (2010). Understanding differences in health behaviorsby education. Journal of Health Economics, 29:1–28.

Edmonds, E. (2002). Reconsidering the labeling effect for child benefits: evidence from atransition economy. Economics Letters, 76(3):303–309.

Frick, J. R. and Groh-Samberg, O. (2007). To claim or not to claim: Estimating non-take-upof social assistance in germany. Discussion Paper 734, DIW, Berlin.

Gohlmann, S. (2007). The determinants of smoking initiation - empirical evidence for Ger-many. Discussion Paper 62, DIW, Berlin.

48

Goux, D. and Maurin, E. (2005). The effect of overcrowded housing on children’s performanceat school. Journal of Public Economics, 89(5-6):797–819.

Green, R. K. and White, M. J. (1997). Measuring the benefits of homeowning: Effects onchildren. Journal of Urban Economics, 41(3):441–461.

Kenkel, D., Schmeiser, M., and Urban, C. (2011). Is smoking inferior? Evidence fromvariation in the earned income tax credit. Working paper, SSRN.

Kooreman, P. (2000). The labeling effect of a child benefit system. American EconomicReview, 90(3):571–583.

Lundberg, S. and Pollack, R. A. (1993). Separate spheres bargaining and the marriagemarket. Journal of Political Economy, 101(6):988–1010.

Lundberg, S. J., Pollack, R. A., and Wales, T. J. (1997). Do husbands and wives pooltheir resources? Evidence from the United Kingdom child benefit. The Journal of HumanResources, 32(3):463–480.

Mabli, J., Castner, L., Ohls, J., Fox, M. K., Crepinsek, M. K., and Condon, E. (2010). Foodexpenditures and diet quality among low-income households and individuals. Technicalreport, Mathematica Policy Research, Washington, DC.

McNown, R. and Ridao-Cano, C. (2004). The effect of child benefit policies on fertilityand female labor force participation in canada. Review of Economics of the Household,2(3):237–254.

Moffitt, R. (1983). An economic model of welfare stigma. The American Economic Review,73(5):10231035.

Moffitt, R. (2005). Remarks on the analysis of causal relationships in population research.Demography, 42(1):91–108.

Nocon, M., Keil, T., and Willich, S. N. (2007). Education, income, occupational status andhealth risk behaviour. Journal of Public Health, 15(5):401–405.

Nord, M., Coleman-Jensen, A., Andrews, M., and Carlson, S. (2010). Household food secu-rity in the United States, 2009. Economic Research Report 108, United States Departmentof Agriculture, Economic Research Service, Washington, DC.

Ruhl, K.-J. (1994). Verordnete Unterordnung: Beruftstatige Frauen zwischenWirtschaftswachstum und konservativer Ideologie in der Nachkriegszeit. R. OldenbourgVerlag, Munchen.

Ruhm, C. J. (2000). Are recessions good for your health? The Quarterly Journal ofEconomics, 115(2):617–650.

Ruhm, C. J. (2005). Healthy living in hard times. Journal of Health Economics, 24(2):341–363.

49

Schady, N. and Rosero, J. (2008). Are cash transfers made to women spent like other sourcesof income? Economics Letters, 101(3):246–248.

Tamm, M. (2008). Does money buy higher schooling? Economics of Education Review,27(5):536–545.

Tamm, M. (2009). Child benefit reform and labor market participation. Ruhr EconomicPapers (Working Paper) 97, Rheinisch-Westfalisches Institut fur Wirtschaftsforschung(RWI), Essen, Germany.

Thaler, R. H. (1990). Anomalies: Saving, fungibility, and mental accounts. Journal ofEconomic Perspectives, 4(1):193–205.

Thaler, R. H. (1999). Mental accounting matters. Journal of Behavioral Decision Making,12:183–206.

Wagner, G. G., Frick, J. R., and Schupp, J. (2007). The German Socio-Economic Panel study(SOEP) - Scope, evolution and enhancements. Schmoller’s Jahrbuch, 127(1):139–169.

Wilkins, N., Yurekli, A., and Hu, T.-W. (2001). Economic analysis of tobacco demand.Technical report, World Bank, Washington, DC.

50

A Appendix Tables

Appendix-Table 1: Robustness Checks of the Impact of Child Benefit on Per CapitaMonthly Food Expenditures in Households.

Per Person Real Monthly Household Food Expenditures

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Contemporaneous Child Benefit and Household Income (Equation 5)

Child Benefit 0.3616∗∗∗ 0.1306∗∗∗ 0.1664∗∗∗ 0.1612∗∗∗ 0.1972∗∗∗ 0.2222∗∗∗

(0.0927) (0.0374) (0.0333) (0.0310) (0.0285) (0.0260)

HH Income 0.0111∗∗∗ 0.0056∗∗∗ 0.0089∗∗∗ 0.0054∗∗∗ 0.0081∗∗∗ 0.0077∗∗∗

(0.0015) (0.0015) (0.0010) (0.0012) (0.0010) (0.0009)

p(CB = HH Income) 0.0002 0.0009 0.0000 0.0000 0.0000 0.0000Observations 8,980 7,987 16,967 10,255 19,235 19,880

Reduced Form (Equation 7)

Child Benefit 0.4004∗∗∗ 0.1421∗∗∗ 0.1853∗∗∗ 0.1708∗∗∗ 0.2119∗∗∗ 0.2309∗∗∗

(0.0925) (0.0369) (0.0330) (0.0307) (0.0282) (0.0255)

Observations 9,294 8,298 17,592 10,634 19,928 20,590

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05, ∗∗∗p < 0.01.p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coefficient of ChildBenefit (CB) equals the coefficient of other household income. Households with young children have atleast one child under the age of 18 living in the households. Families with older children have childrenwho are all at least 18 years of age. Household income does not include the Child Benefit. Both ChildBenefit and Household Income are Measured in Real 2005 e. Household income is lagged by one year.Other controls included in each specification are single parent indicators, a quadratic term in the age ofthe youngest child, the state level unemployment rate, an dummy variables for whether the country wasin a recession during a sample year, dummy variables for states as well as state specific year trends. Spec-ifications for Columns (3)-(6) also control for the number of children in the household.

51

Appendix-Table 2: Robustness Check:The Impact of Child Benefit on Per CapitaMonthly Food Expenditures in Households using a log specifica-tion.

Log Per Person Real Monthly Household Food Expenditures

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Households with children of any age

Child Benefit 0.0025∗∗∗ 0.0008∗∗ 0.0010∗∗∗ 0.0011∗∗∗ 0.0012∗∗∗ 0.0010∗∗∗

(0.0006) (0.0003) (0.0002) (0.0002) (0.0002) (0.0002)

HH Income 0.0000∗∗∗ 0.0000∗ 0.0000∗∗∗ 0.0000∗ 0.0000∗∗∗ 0.0000∗∗∗

(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)p(CB = HH Income) 0.0000 0.0031 0.0000 0.0000 0.0000 0.0000Observations 6,942 6,856 13,798 8,803 15,745 16,295

Households with young children

Child Benefit 0.0030∗∗∗ 0.0009∗∗ 0.0011∗∗∗ 0.0011∗∗∗ 0.0012∗∗∗ 0.0010∗∗∗

(0.0007) (0.0003) (0.0002) (0.0002) (0.0002) (0.0002)

HH Income 0.0000∗∗ 0.0000∗∗ 0.0000∗∗∗ 0.0000∗∗ 0.0000∗∗∗ 0.0000∗∗∗

(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)p(CB = HH Income) 0.0001 0.0029 0.0000 0.0000 0.0000 0.0000Observations 5,067 6,292 11,359 8,182 13,249 13,793

Households with old children

Child Benefit 0.0026 0.0024 0.0017 0.0011 0.0016 0.0013(0.0018) (0.0022) (0.0012) (0.0019) (0.0011) (0.0011)

HH Income 0.0000 -0.0000 0.0000 -0.0000 0.0000 0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.1565 0.2749 0.1388 0.5624 0.1554 0.2362Observations 1,406 355 1,761 391 1,797 1,800

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05,∗∗∗p < 0.01. p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coef-ficient of Child Benefit (CB) equals the coefficient of other household income. Households with youngchildren have at least one child under the age of 18 living in the households. Families with older childrenhave children who are all at least 18 years of age. Household income does not include the Child Benefit.Both Child Benefit and Household Income are Measured in Real 2005 e. Household income is lagged byone year. Other controls included in each specification are single parent indicators, a quadratic term inthe age of the youngest child, the state level unemployment rate, an dummy variables for whether thecountry was in a recession during a sample year, dummy variables for states as well as state specific yeartrends. Specifications for Columns (3)-(6) also control for the number of children in the household.

52

Appendix-Table 3: Robustness Checks of the Impact of Child Benefit on the Probabilitythat Households Rent their Home.

Probability of Being a Renter

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Contemporaneous Child Benefit and Household Income (Equation 5)

Child Benefit -0.0001∗∗ -0.0001∗∗ -0.0002∗∗∗ -0.0001∗∗ -0.0001∗∗∗ -0.0001∗∗∗

(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

HH Income -0.0000∗∗∗ -0.0000∗∗∗ -0.0000∗∗∗ -0.0000∗∗∗ -0.0000∗∗∗ -0.0000∗∗∗

(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.0054 0.0284 0.0000 0.0134 0.0000 0.0000Observations 13,191 12,871 26,062 16,459 29,650 30,617

Reduced Form (Equation 7)

Child Benefit -0.0001∗∗ -0.0001∗ -0.0001∗∗∗ -0.0001∗∗ -0.0001∗∗∗ -0.0001∗∗∗

(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

Observations 13,654 13,389 27,043 17,079 30,733 31,718

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05, ∗∗∗p < 0.01.p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coefficient of Child Ben-efit (CB) equals the coefficient of other household income. Households with young children have at least onechild under the age of 18 living in the households. Families with older children have children who are all atleast 18 years of age. Household income does not include the Child Benefit. Both Child Benefit and House-hold Income are Measured in Real 2005 e. Child Benefit is lagged by one year. Household income is laggedby two years. Other controls included in each specification are single parent indicators, a quadratic term inthe age of the youngest child, the state level unemployment rate, an dummy variables for whether the coun-try was in a recession during a sample year, dummy variables for states as well as state specific year trends.Specifications for Columns (3)-(6) also control for the number of children in the household.

53

Appendix-Table 4: Robustness Checks of the Impact of Child Benefit on the Size(Square Meters) of Homes of Households.

Size of Home in Square Meters

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Contemporaneous Child Benefit and Household Income (Equation 5)

Child Benefit 0.0092∗ 0.0094∗ 0.0153∗∗∗ 0.0109∗∗ 0.0161∗∗∗ 0.0150∗∗∗

(0.0046) (0.0043) (0.0036) (0.0040) (0.0034) (0.0032)

HH Income 0.0014∗∗∗ 0.0008 0.0014∗∗∗ 0.0011∗∗ 0.0014∗∗∗ 0.0015∗∗∗

(0.0003) (0.0005) (0.0003) (0.0004) (0.0003) (0.0003)

p(CB = HH Income) 0.0858 0.0473 0.0001 0.0159 0.0000 0.0000Observations 13,191 12,871 26,062 16,459 29,650 30,617

Reduced Form (Equation 7)

Child Benefit 0.0070 0.0087∗ 0.0138∗∗∗ 0.0104∗∗ 0.0150∗∗∗ 0.0144∗∗∗

(0.0042) (0.0042) (0.0035) (0.0040) (0.0033) (0.0031)

Observations 13,654 13,389 27,043 17,079 30,733 31,718

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05,∗∗∗p < 0.01. p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coef-ficient of Child Benefit (CB) equals the coefficient of other household income. Households with youngchildren have at least one child under the age of 18 living in the households. Families with older childrenhave children who are all at least 18 years of age. Household income does not include the Child Benefit.Both Child Benefit and Household Income are Measured in Real 2005 e. Child Benefit is lagged by oneyear. Household income is lagged by two years. Other controls included in each specification are singleparent indicators, a quadratic term in the age of the youngest child, the state level unemployment rate,an dummy variables for whether the country was in a recession during a sample year, dummy variablesfor states as well as state specific year trends. Specifications for Columns (3)-(6) also control for thenumber of children in the household.

54

Appendix-Table 5: Robustness Checks of the Impact of Child Benefit on the Numberof Rooms in the Homes of Households.

Number of Rooms

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Contemporaneous Child Benefit and Household Income (Equation 5)

Child Benefit 0.0004 0.0004∗∗ 0.0007∗∗∗ 0.0005∗∗∗ 0.0007∗∗∗ 0.0007∗∗∗

(0.0002) (0.0002) (0.0001) (0.0001) (0.0001) (0.0001)

HH Income 0.0000∗∗ 0.0000∗∗ 0.0000∗∗∗ 0.0000∗∗ 0.0000∗∗∗ 0.0000∗∗∗

(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.0853 0.0084 0.0000 0.0019 0.0000 0.0000Observations 13,191 12,871 26,062 16,459 29,650 30,617

Reduced Form (Equation 7)

Child Benefit 0.0002 0.0005∗∗ 0.0006∗∗∗ 0.0005∗∗∗ 0.0007∗∗∗ 0.0007∗∗∗

(0.0002) (0.0002) (0.0001) (0.0001) (0.0001) (0.0001)

Observations 13,654 13,389 27,043 17,079 30,733 31,718

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05,∗∗∗p < 0.01. p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coef-ficient of Child Benefit (CB) equals the coefficient of other household income. Households with youngchildren have at least one child under the age of 18 living in the households. Families with older childrenhave children who are all at least 18 years of age. Household income does not include the Child Benefit.Both Child Benefit and Household Income are Measured in Real 2005 e. Child Benefit is lagged by oneyear. Household income is lagged by two years. Other controls included in each specification are singleparent indicators, a quadratic term in the age of the youngest child, the state level unemployment rate,an dummy variables for whether the country was in a recession during a sample year, dummy variablesfor states as well as state specific year trends. Specifications for Columns (3)-(6) also control for thenumber of children in the household.

55

Appendix-Table 6: Robustness Checks for the Impact of Child Benefit on the Prob-ability that Parents Smoke.

Smoking Participation

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Contemporaneous Child Benefit and Household Income (Equation 5)

Child Benefit 0.0003 0.0002 0.0002 -0.0000 -0.0000 0.0000(0.0005) (0.0002) (0.0002) (0.0002) (0.0002) (0.0001)

HH Income 0.0000∗ -0.0000 0.0000 -0.0000 0.0000 0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.6272 0.2834 0.2752 0.9199 0.9953 0.7319Observations 15,131 14,014 29,145 18,074 33,205 34,329

Reduced Form (Equation 7)

Child Benefit 0.0002 0.0002 0.0002 -0.0001 -0.0000 -0.0000(0.0005) (0.0002) (0.0002) (0.0002) (0.0002) (0.0001)

Observations 15,718 14,643 30,361 18,824 34,542 35,693

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05,∗∗∗p < 0.01. p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coef-ficient of Child Benefit (CB) equals the coefficient of other household income. Households with youngchildren have at least one child under the age of 18 living in the households. Families with older childrenhave children who are all at least 18 years of age. Household income does not include the Child Benefit.Both Child Benefit and Household Income are Measured in Real 2005 e. Household income is laggedby one year. Other controls included in each specification are single parent indicators, a quadratic termin the age of the youngest child, the state level unemployment rate, an dummy variables for whetherthe country was in a recession during a sample year, dummy variables for states as well as state specificyear trends. Specifications for Columns (3)-(6) also control for the number of children in the household.

56

Appendix-Table 7: Robustness Checks of the Impact of Child Benefit on the Proba-bility that Parents Drink Alcohol Regularly.

Drinking Alcohol Regularly

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Contemporaneous Child Benefit and Household Income (Equation 5)

Child Benefit -0.0664 0.0362 -0.0004 0.0039 0.0012 0.0007(0.0911) (0.0725) (0.0017) (0.0026) (0.0012) (0.0009)

HH Income -0.0000 0.0000 0.0000 0.0000 0.0000 0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.4665 0.6174 0.8026 0.1413 0.3131 0.4083Observations 4,789 4,395 9,184 5,624 10,413 10,727

Reduced Form (Equation 7)

Child Benefit -0.0679 0.0361 -0.0012 0.0032 0.0005 0.0004(0.0857) (0.0719) (0.0016) (0.0026) (0.0012) (0.0008)

Observations 4,948 4,619 9,567 5,894 10,842 11,159

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05,∗∗∗p < 0.01. p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coef-ficient of Child Benefit (CB) equals the coefficient of other household income. Households with youngchildren have at least one child under the age of 18 living in the households. Families with older childrenhave children who are all at least 18 years of age. Household income does not include the Child Benefit.Both Child Benefit and Household Income are Measured in Real 2005 e. Household income is laggedby one year. Other controls included in each specification are single parent indicators, a quadratic termin the age of the youngest child, the state level unemployment rate, an dummy variables for whetherthe country was in a recession during a sample year, dummy variables for states as well as state specificyear trends. Specifications for Columns (3)-(6) also control for the number of children in the household.

57

Appendix-Table 8: Robustness Checks of the Impact of Child Benefit on the Probabilitythat Parents Go to the Movies or Attend Pop Music Concerts atleast Once per Month.

Going to the Movies or Concerts Once a Month

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Contemporaneous Child Benefit and Household Income (Equation 5)

Child Benefit 0.0028∗∗∗ 0.0011∗∗∗ 0.0011∗∗∗ 0.0016∗∗∗ 0.0014∗∗∗ 0.0014∗∗∗

(0.0004) (0.0002) (0.0002) (0.0002) (0.0001) (0.0001)

HH Income -0.0000 0.0000 0.0000 0.0000∗ 0.0000 0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000Observations 18,834 17,368 36,202 22,305 41,139 42,512

Reduced Form (Equation 7)

Child Benefit 0.0027∗∗∗ 0.0011∗∗∗ 0.0011∗∗∗ 0.0016∗∗∗ 0.0014∗∗∗ 0.0014∗∗∗

(0.0004) (0.0002) (0.0002) (0.0002) (0.0001) (0.0001)

Observations 19,529 18,062 37,591 23,158 42,687 44,089

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05, ∗∗∗p < 0.01.p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coefficient of ChildBenefit (CB) equals the coefficient of other household income. Households with young children have atleast one child under the age of 18 living in the households. Families with older children have childrenwho are all at least 18 years of age. Household income does not include the Child Benefit. Both ChildBenefit and Household Income are Measured in Real 2005 e. Household income is lagged by one year.Other controls included in each specification are single parent indicators, a quadratic term in the age ofthe youngest child, the state level unemployment rate, an dummy variables for whether the country wasin a recession during a sample year, dummy variables for states as well as state specific year trends. Spec-ifications for Columns (3)-(6) also control for the number of children in the household.

58

Appendix-Table 9: Robustness Checks of the Impact of Child Benefit on the Probabilitythat Parents Go to the Opera or to Public Lectures at least onceper Month.

Going to the Opera or Attending Public Lectures Once a Month

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Contemporaneous Child Benefit and Household Income (Equation 5)

Child Benefit 0.0025∗∗∗ 0.0010∗∗∗ 0.0010∗∗∗ 0.0015∗∗∗ 0.0013∗∗∗ 0.0012∗∗∗

(0.0003) (0.0001) (0.0001) (0.0001) (0.0001) (0.0001)

HH Income 0.0000 0.0000 0.0000∗∗ 0.0000∗ 0.0000∗∗∗ 0.0000∗∗∗

(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000Observations 18,848 17,373 36,221 22,311 41,159 42,533

Households with young children (Equation 7)

Child Benefit 0.0026∗∗∗ 0.0011∗∗∗ 0.0011∗∗∗ 0.0015∗∗∗ 0.0014∗∗∗ 0.0013∗∗∗

(0.0003) (0.0001) (0.0001) (0.0001) (0.0001) (0.0001)

Observations 19,546 18,067 37,613 23,162 42,708 44,111

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05, ∗∗∗p < 0.01.p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coefficient of ChildBenefit (CB) equals the coefficient of other household income. Households with young children have atleast one child under the age of 18 living in the households. Families with older children have childrenwho are all at least 18 years of age. Household income does not include the Child Benefit. Both ChildBenefit and Household Income are Measured in Real 2005 e. Household income is lagged by one year.Other controls included in each specification are single parent indicators, a quadratic term in the age ofthe youngest child, the state level unemployment rate, an dummy variables for whether the country wasin a recession during a sample year, dummy variables for states as well as state specific year trends. Spec-ifications for Columns (3)-(6) also control for the number of children in the household.

59

Appendix-Table 10: Robustness Checks of the Impact of Child Benefit on the Prob-ability that Parents Make Excursions or Go on Short VacationTrips at least Once per Month.

Making Excursions or Going on Vacation Trips Once a Month

(1) (2) (3) (4) (5) (6)

Families withAny

1 2 1 or 2 2 or 3 1, 2, or 3 Number ofChild Children Children Children Children Children

Contemporaneous Child Benefit and Household Income (Equation 5)

Child Benefit 0.0012 0.0007 0.0010∗∗ 0.0006 0.0006∗ 0.0005∗

(0.0009) (0.0004) (0.0003) (0.0003) (0.0003) (0.0002)

HH Income 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

p(CB = HH Income) 0.1936 0.0813 0.0019 0.0759 0.0263 0.0528Observations 6,951 6,387 13,338 8,221 15,172 15,670

Reduced Form (Equation 7)

Child Benefit 0.0016 0.0006 0.0009∗∗ 0.0005 0.0006∗ 0.0004(0.0009) (0.0004) (0.0003) (0.0003) (0.0003) (0.0002)

Observations 7,212 6,650 13,862 8,550 15,762 16,269

Note: Standard errors clustered at the household level are in parenthesis. ∗p < 0.1, ∗∗p < 0.05,∗∗∗p < 0.01. p(CB = HH Income) is the p-value of an F-test testing the null hypothesis that the coef-ficient of Child Benefit (CB) equals the coefficient of other household income. Households with youngchildren have at least one child under the age of 18 living in the households. Families with older childrenhave children who are all at least 18 years of age. Household income does not include the Child Benefit.Both Child Benefit and Household Income are Measured in Real 2005 e. Household income is laggedby one year. Other controls included in each specification are single parent indicators, a quadratic termin the age of the youngest child, the state level unemployment rate, an dummy variables for whetherthe country was in a recession during a sample year, dummy variables for states as well as state specificyear trends. Specifications for Columns (3)-(6) also control for the number of children in the household.

60